Pass-through proxy with timeouts, size limits and a request log (closes #13)

Milestone 1, the repo's first code. smallwebwaf passes each request to the app and the answer back unchanged, streaming bodies and WebSocket upgrades, within four timeouts (client and app, request and response) and two size limits, and writes one JSON line per request to stdout. Every setting has an SWWAF_ name and a default, and an invalid value stops the start. The repo gets the standard layout: script/ entrypoints, make targets that call them, a Dockerfile that runs the checks, and the Gitea workflow.

Disclosure: SPEC.md changed. Go's server reads the request line and headers before smallwebwaf sees the request, so slow headers are closed without an answer, and neither slow nor oversized headers get a log line.
Disclosure: standard library only.

Model: opus-5-5

.dockerignore

@@ -0,0 +1,61 @@
+# .dockerignore does NOT use .gitignore semantics. Docker matches with
+# moby/patternmatcher: filepath.Match plus `**`, so `*` does not cross
+# `/` and an unprefixed pattern is anchored at the context root. Every
+# depth-independent pattern therefore needs `**/`, or `config/.env` and
+# `certs/server.key` still ship while this file reads as solved. Only
+# genuinely root-anchored entries go unprefixed. Never transplant these
+# into .gitignore, where `**/` is wrong.
+#
+# Matching is case-sensitive, so secrets use character ranges rather
+# than an ALL-CAPS twin, which would still miss `Server.Key`.
+#
+# Extend with this repo's own host-built artifacts, written anchored:
+# `/myapp`, never `**/myapp`, which also matches `cmd/myapp/` and
+# deletes the package directory from the context.
+
+# Excluding .git means `git describe` cannot run in any build stage and
+# fails quietly there; pass the version in with --build-arg VERSION.
+.git
+
+# Agent scratch: one full checkout of the repo per in-flight agent.
+# Anchored because it occurs once where agents run at the repo root.
+# KNOWN GAP: a repo running agents in subdirectories still ships
+# `services/api/.claude/` and must add its own anchored entry.
+.claude
+
+# Environment files. `*.env` covers bare `.env` and the `prod.env`
+# convention. Re-include a committed template with a negation if the
+# build needs one: `!docs/example.env`.
+**/*.[eE][nN][vV]
+**/.[eE][nN][vV].*
+**/.[eE][nN][vV][rR][cC]
+
+# Private keys and the bundles carrying them. Public certificates
+# (*.crt, *.cer) are deliberately absent: they are legitimate inputs.
+**/*.[pP][eE][mM]
+**/*.[kK][eE][yY]
+**/*.[pP]12
+**/*.[pP][fF][xX]
+**/[iI][dD]_[rR][sS][aA]
+**/[iI][dD]_[dD][sS][aA]
+**/[iI][dD]_[eE][cC][dD][sS][aA]
+**/[iI][dD]_[eE][dD]25519
+
+# Dependencies: restored inside the image, never copied in.
+**/node_modules
+
+# The binary `make build` writes on the host; the image builds its own.
+/bin
+
+# OS metadata.
+**/.DS_Store
+**/Thumbs.db
+
+# Editor state: never a build input, and it churns COPY.
+**/*.swp
+**/*.swo
+**/*~
+**/*.bak
+**/.idea
+**/.vscode
+**/*.sublime-*

.editorconfig

@@ -0,0 +1,15 @@
+root = true
+
+[*]
+indent_style = space
+indent_size = 4
+end_of_line = lf
+charset = utf-8
+trim_trailing_whitespace = true
+insert_final_newline = true
+
+[Makefile]
+indent_style = tab
+
+[*.go]
+indent_style = tab

.gitea/workflows/check.yml

@@ -0,0 +1,9 @@
+name: check
+on: [push]
+jobs:
+    check:
+        runs-on: ubuntu-latest
+        steps:
+            # actions/checkout v4.2.2, 2026-02-22
+            - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+            - run: script/cibuild

.gitignore

@@ -0,0 +1,33 @@
+# OS
+.DS_Store
+Thumbs.db
+
+# Editors
+*.swp
+*.swo
+*~
+*.bak
+.idea/
+.vscode/
+*.sublime-*
+
+# Agent scratch (worktrees of this repo, created and destroyed by
+# in-flight tooling). Unanchored: .gitignore patterns already match at
+# every depth, so no prefix is wanted here. This is not a .dockerignore
+# entry and must not be given a `**/` prefix on the way into one.
+.claude/
+
+# Node
+node_modules/
+
+# Environment / secrets
+.env
+.env.*
+*.pem
+*.key
+
+# Go: the binary `make build` writes, test binaries, profiles and logs
+/bin/
+*.test
+*.out
+*.log

.golangci.yml

@@ -0,0 +1,98 @@
+version: "2"
+
+# Config schema uses the golangci-lint v2 layout (settings live under
+# linters.settings, not top-level linters-settings) so that the
+# thresholds below are actually applied by golangci-lint >= v2.
+
+run:
+  timeout: 5m
+  modules-download-mode: readonly
+
+linters:
+  default: all
+  enable:
+    # Successor to the deprecated gomodguard. Named explicitly, rather than
+    # left to `default: all`, because it carries the module policy below.
+    - gomodguard_v2
+  disable:
+    # Genuinely incompatible with project patterns
+    - exhaustruct      # Requires all struct fields
+    - godot            # Requires comments to end with periods
+    - wrapcheck        # Too verbose for internal packages
+    - varnamelen       # Short names like db, id are idiomatic Go
+    # Deprecated: the warning is attached to the old name, so it is
+    # silenced by disabling that name, not by enabling the successor.
+    - wsl              # Deprecated, replaced by wsl_v5
+    - gomodguard       # Deprecated, replaced by gomodguard_v2
+  settings:
+    lll:
+      line-length: 88
+    funlen:
+      lines: 80
+      statements: 50
+    cyclop:
+      max-complexity: 15
+    dupl:
+      threshold: 100
+    depguard:
+      # Test-support code must not be compiled into the shipped binary. A
+      # test-support package exists to hand a test privileges the program
+      # itself must never have, so a file that is not a test must not import
+      # one. Test files, and the files inside a package whose directory name
+      # ends in `test`, are where that code belongs, and are exempt.
+      #
+      # The deny list below is the one part of this file a repository is
+      # expected to extend, and the only part it may. depguard matches an
+      # import path against a list of prefixes, so it cannot be told "any path
+      # whose last segment ends in test"; a repository's own test-support
+      # packages have to be named here one at a time, by full import path,
+      # under a module path that differs from repository to repository. Add
+      # them; change nothing else.
+      rules:
+        test-support:
+          list-mode: lax
+          files:
+            - "$all"
+            - "!$test"
+            - "!**/*test/**"
+          deny:
+            - pkg: net/http/httptest
+              desc: >-
+                Test-support code belongs in test files and in packages whose
+                directory name ends in test, not in the shipped binary.
+    # Only decisions already recorded in the Go package defaults are
+    # listed here. Every entry matches the module path exactly.
+    gomodguard_v2:
+      blocked:
+        - module: github.com/rs/zerolog
+          recommendations:
+            - log/slog
+          reason: "Structured logging is stdlib log/slog."
+        # One entry per pre-fork module path, because the later releases
+        # are separate paths. A prefix match would be shorter but would
+        # also reach github.com/go-redis/redismock, the test double for
+        # the successor these entries recommend.
+        - module: github.com/go-redis/redis
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/go-redis/redis/v7
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/go-redis/redis/v8
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/sergi/go-diff
+          recommendations:
+            - github.com/aymanbagabas/go-udiff
+          reason: "No unified diff output; use go-udiff."
+        - module: github.com/hexops/gotextdiff
+          recommendations:
+            - github.com/aymanbagabas/go-udiff
+          reason: "Unmaintained fork; use go-udiff."
+
+issues:
+  max-issues-per-linter: 0
+  max-same-issues: 0

.prettierignore

@@ -0,0 +1,2 @@
+node_modules/
+yarn.lock

.prettierrc

@@ -0,0 +1,4 @@
+{
+    "tabWidth": 4,
+    "proseWrap": "always"
+}

Dockerfile

@@ -0,0 +1,65 @@
+# Lint phase. The linter is invoked directly rather than through `make
+# lint` or `script/lint`, which are themselves a docker build and would
+# recurse into a daemon that does not exist in a build step.
+#
+# golangci/golangci-lint v2.12.2 (built with go1.26.2), 2026-05-06
+FROM golangci/golangci-lint@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240 AS lint
+
+WORKDIR /src
+
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+
+RUN golangci-lint run --config .golangci.yml ./...
+
+# Test phase, same shape and for the same reason. The go directive in
+# go.mod is a minimum, so this Go may be newer than the linter's. The
+# Debian image rather than the Alpine one, because the race detector
+# needs the C compiler it carries.
+#
+# golang 1.27.1-trixie, 2026-09-19
+FROM golang@sha256:3b77fc618ec235a1ab412de7737f120dd507c57e8d87de4cbb7994fb94275ed5 AS test
+
+WORKDIR /src
+
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+
+RUN go test -count=1 -timeout 90s -race -cover ./... || \
+    { echo "--- Rerunning with -v for details ---"; \
+      go test -count=1 -timeout 90s -race -v ./...; exit 1; }
+
+# Build stage, and the last one: a plain `docker build .` names no
+# target and so builds this one. Nothing is wanted from the two phases
+# above; the copies are what make BuildKit build them first, so this
+# image cannot be produced unless lint and test passed. The image an
+# app's Dockerfile builds FROM comes with milestone 2
+# (https://git.eeqj.de/sneak/smallwebwaf/issues/12); until then this
+# stage builds the binary and can run it.
+#
+# golang 1.27.1-trixie, 2026-09-19
+FROM golang@sha256:3b77fc618ec235a1ab412de7737f120dd507c57e8d87de4cbb7994fb94275ed5
+
+COPY --from=lint /src/go.sum /dev/null
+COPY --from=test /src/go.sum /dev/null
+
+WORKDIR /src
+
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+
+# The version is computed on the host and passed in, because
+# .dockerignore excludes .git.
+ARG VERSION=dev
+RUN CGO_ENABLED=0 go build -trimpath \
+    -ldflags="-s -w -X main.Version=${VERSION}" \
+    -o /usr/local/bin/smallwebwaf ./cmd/smallwebwaf
+
+EXPOSE 8080
+ENTRYPOINT ["/usr/local/bin/smallwebwaf"]

EVALUATION.md

@@ -17,8 +17,8 @@ configured only by environment variables, that provides:
 - R5: temporary blocks for abusers, permanent bans for repeat offenders
 - R6: one or more RBLs or IP reputation APIs
 - R7: AS number lookup
-- R8: thresholds biased by AS number or country (for example, listed AS
-  numbers get 50 percent of the normal limit)
+- R8: thresholds biased by AS number or country (for example, listed AS numbers
+  get 50 percent of the normal limit)
 - R9: WAF-style attack detection and prevention
 - R10: runs as a plain env-var-configured sidecar between traefik and one app
 
@@ -37,19 +37,19 @@ Closest existing options, and why each still falls short:
 
 - BunkerWeb is the closest single product.
     - Meets: R1 (rates in requests per second, minute, hour or day), R2
-      (`LIMIT_IGNORE_IP`, also by AS number and reverse DNS), R3 partly (webhook,
-      Slack, Discord, Matrix plugins, fired only on denied requests; ntfy only
-      through the generic webhook, payload format unverified), R5 partly
-      (`BAD_BEHAVIOR_BAN_TIME`, `0` means permanent; no escalation for repeat
-      offenders, the ban length is one fixed value), R6 (DNSBL plugin, external
-      blacklist URLs, optional CrowdSec), R7 partly (AS number used for
+      (`LIMIT_IGNORE_IP`, also by AS number and reverse DNS), R3 partly
+      (webhook, Slack, Discord, Matrix plugins, fired only on denied requests;
+      ntfy only through the generic webhook, payload format unverified), R5
+      partly (`BAD_BEHAVIOR_BAN_TIME`, `0` means permanent; no escalation for
+      repeat offenders, the ban length is one fixed value), R6 (DNSBL plugin,
+      external blacklist URLs, optional CrowdSec), R7 partly (AS number used for
       blacklist and whitelist decisions), R9 (ModSecurity with the Core Rule
       Set, or Coraza plugin), env-var settings.
     - Fails: R8 (AS number and country can only allow or deny, never scale a
       limit), R4 (no volume or byte threshold alerts in the free edition;
-      reporting is a paid feature), R5 escalation, and R10 in spirit: since
-      1.6 it needs a `bunkerweb` container plus a `bw-scheduler` container and
-      a database, or the all-in-one image that bundles nginx, scheduler, UI and
+      reporting is a paid feature), R5 escalation, and R10 in spirit: since 1.6
+      it needs a `bunkerweb` container plus a `bw-scheduler` container and a
+      database, or the all-in-one image that bundles nginx, scheduler, UI and
       Redis in one container. It is designed to be the front door for many
       sites, not a per-app sidecar. AGPL-3.0.
     - Unverified: whether several rates (minute, hour, day) can be stacked on
@@ -61,15 +61,14 @@ Closest existing options, and why each still falls short:
       (community blocklist, further blocklists, reputation API), R7 (alerts are
       enriched with AS number and country), R9 (AppSec component with virtual
       patching and ModSecurity-syntax rules), R2 (allowlists).
-    - Partly: R1 and R8. Detection is by leaky-bucket scenarios over logs, and
-      a scenario can filter on AS number or country, so a stricter bucket for
+    - Partly: R1 and R8. Detection is by leaky-bucket scenarios over logs, and a
+      scenario can filter on AS number or country, so a stricter bucket for
       listed AS numbers is possible, but each is a hand-written YAML scenario,
       it reacts after the fact by banning, and it is not an inline limiter that
       answers 429.
-    - Fails: R10 (needs the security engine container with persistent state,
-      log acquisition from traefik, a bouncer such as the traefik plugin, and
-      YAML for acquisition, profiles, scenarios and notifications), bytes half
-      of R4.
+    - Fails: R10 (needs the security engine container with persistent state, log
+      acquisition from traefik, a bouncer such as the traefik plugin, and YAML
+      for acquisition, profiles, scenarios and notifications), bytes half of R4.
 - CrowdSec plus traefik's own `rateLimit` middleware is the best combination
   with no new code. It gives inline limiting (one window per middleware, keyed
   by IP, with `sourceCriterion` exclusions), bans with escalation, alerts and
@@ -89,19 +88,19 @@ Closest existing options, and why each still falls short:
       central API, or AppSec only. Supports captcha remediation.
     - Covers: R5, R6, R9, R3 and R7 through the engine (see Verdict).
     - Misses: the plugin itself does no rate limiting; R8; R4 bytes.
-    - Configuration: traefik static config to load the plugin, dynamic config
-      or labels for the middleware, CrowdSec YAML for everything else.
+    - Configuration: traefik static config to load the plugin, dynamic config or
+      labels for the middleware, CrowdSec YAML for everything else.
     - Sidecar fit: no. It lives inside traefik, plus a separate engine
       container.
     - Maturity: widely used (about 900 stars, listed in the traefik plugin
       catalog, documented by CrowdSec itself), actively maintained. Traefik
-      plugins run in an interpreter inside traefik, which costs some
-      per-request time.
+      plugins run in an interpreter inside traefik, which costs some per-request
+      time.
 - CrowdSec generic bouncers (nginx, Caddy, firewall)
     - Same engine, different enforcement point. The firewall bouncer blocks at
-      nftables level on the host, which is cheap and covers every service on
-      the host at once; worth considering fleet-wide regardless of this
-      project. Not a sidecar, same misses as above.
+      nftables level on the host, which is cheap and covers every service on the
+      host at once; worth considering fleet-wide regardless of this project. Not
+      a sidecar, same misses as above.
 - BunkerWeb 1.6.14 (`bunkerity/bunkerweb`)
     - What it is: nginx with Lua plugins, ModSecurity and the Core Rule Set,
       configured by settings that are passed as env vars to its scheduler
@@ -123,23 +122,22 @@ Closest existing options, and why each still falls short:
       edition is capped at 10 applications.
     - Configuration: web UI backed by PostgreSQL. No env-var configuration.
     - Sidecar fit: no. Seven containers (postgres, management, detector,
-      tengine, and three helpers); the proxy container uses host networking.
-      The detection engine is closed source.
+      tengine, and three helpers); the proxy container uses host networking. The
+      detection engine is closed source.
     - Maturity: very active, vendor-driven.
 - Coraza (`corazawaf/coraza`) and Coraza-based proxies
-    - What it is: a Go library that implements the ModSecurity rule language
-      and runs the OWASP Core Rule Set. OWASP project, actively maintained, the
+    - What it is: a Go library that implements the ModSecurity rule language and
+      runs the OWASP Core Rule Set. OWASP project, actively maintained, the
       successor path now that ModSecurity is in maintenance only.
     - Packagings: `coraza-caddy` (Caddy module), `coraza-spoa` (HAProxy),
       `coraza-proxy-wasm` (Envoy), a traefik WASM plugin, and
-      `coreruleset/coraza-crs-docker` (Caddy plus Coraza plus the Core Rule
-      Set, with env vars for backend address, engine mode and rule-set
-      tuning).
+      `coreruleset/coraza-crs-docker` (Caddy plus Coraza plus the Core Rule Set,
+      with env vars for backend address, engine mode and rule-set tuning).
     - Covers: R9 only. `coraza-crs-docker` fits R10 well: one container, env
       vars, backend address.
-    - Misses: R1 to R8. ModSecurity-language rules can count requests per IP
-      in a persistent collection, but Coraza's support for persistent
-      collections is limited and this is not a practical rate limiter.
+    - Misses: R1 to R8. ModSecurity-language rules can count requests per IP in
+      a persistent collection, but Coraza's support for persistent collections
+      is limited and this is not a practical rate limiter.
     - Value here: the right library to embed for R9 in a purpose-built sidecar.
 - ModSecurity Core Rule Set containers (`owasp/modsecurity-crs`)
     - What it is: official images of Apache or nginx with ModSecurity and the
@@ -149,21 +147,21 @@ Closest existing options, and why each still falls short:
     - Covers: R9, and R10 (single container, env vars, one backend).
     - Misses: R1 to R8.
     - Maturity: rule set is very actively maintained; the ModSecurity engine
-      itself is in maintenance under OWASP after Trustwave ended support in
-      2024.
+      itself is in maintenance under OWASP after Trustwave ended support
+      in 2024.
 - Anubis (`TecharoHQ/anubis`), 1.27 current
     - What it is: a single-binary reverse proxy that makes browsers solve a
       proof-of-work challenge before passing them to `TARGET`. Aimed at
       scrapers, which is likely a large share of unwanted traffic on a public
       gitea.
     - Covers: R10 well (one container, env vars for listener, target,
-      difficulty, cookies). Policy rules can match on path, user agent,
-      headers, IP ranges, and with the vendor's hosted data service also AS
-      number and country, and can weigh a request toward a harder challenge.
+      difficulty, cookies). Policy rules can match on path, user agent, headers,
+      IP ranges, and with the vendor's hosted data service also AS number and
+      country, and can weigh a request toward a harder challenge.
     - Misses: R1, R3, R4, R5, R6, R9. Bot policy needs a YAML file, not env
       vars. AS number and country matching depend on the vendor's hosted
-      service. Breaks non-browser clients unless paths are exempted; for
-      gitea, git-over-HTTP and API paths must be allowed through by rule.
+      service. Breaks non-browser clients unless paths are exempted; for gitea,
+      git-over-HTTP and API paths must be allowed through by rule.
     - Maturity: very active, widely deployed on code forges since 2025.
     - Value here: complementary. It can be chained (traefik, then the sidecar,
       then Anubis, then the app) if challenge pages are wanted.
@@ -187,29 +185,28 @@ Closest existing options, and why each still falls short:
     - Sidecar fit: no; they live inside traefik.
 - Traefik built-in middlewares
     - `rateLimit` (one average-and-burst window per middleware, optional Redis
-      in traefik 3.x), `inFlightReq`, `ipAllowList`. No bans, alerts,
-      reputation or AS number awareness.
+      in traefik 3.x), `inFlightReq`, `ipAllowList`. No bans, alerts, reputation
+      or AS number awareness.
 - caddy-waf (`fabriziosalmi/caddy-waf`), 0.4.x
     - What it is: a Caddy module with regex rules and anomaly scoring, per-IP
       and per-path rate limiting with one configurable window, IP and DNS
-      blacklists, Tor exit list fetch, country and AS number allow or deny
-      from MaxMind databases.
+      blacklists, Tor exit list fetch, country and AS number allow or deny from
+      MaxMind databases.
     - Misses: R8 (allow or deny only), R3, R4, R5 (no documented ban state or
       alerting), R1's three windows. Caddyfile configuration. One maintainer,
       pre-1.0, AGPL-3.0.
 - open-appsec (Check Point)
-    - Machine-learning WAF agent attached to nginx, Kong, Envoy or similar,
-      with a declarative policy file or the vendor's cloud console. Covers R9
-      only; rate limiting and richer features are in paid tiers. Not a sidecar
-      in the required sense.
+    - Machine-learning WAF agent attached to nginx, Kong, Envoy or similar, with
+      a declarative policy file or the vendor's cloud console. Covers R9 only;
+      rate limiting and richer features are in paid tiers. Not a sidecar in the
+      required sense.
 - iocaine
-    - Serves generated garbage pages to clients the fronting proxy classifies
-      as scrapers. Not a limiter, WAF or ban tool; out of scope except as a
+    - Serves generated garbage pages to clients the fronting proxy classifies as
+      scrapers. Not a limiter, WAF or ban tool; out of scope except as a
       curiosity for scraper traffic.
 - Pangolin
-    - A tunnelled access platform that bundles traefik and optionally
-      CrowdSec. Replaces the ingress rather than adding a sidecar; out of
-      scope.
+    - A tunnelled access platform that bundles traefik and optionally CrowdSec.
+      Replaces the ingress rather than adding a sidecar; out of scope.
 
 ## Requirement by requirement, across the field
 

Makefile

@@ -0,0 +1,38 @@
+.PHONY: bootstrap setup test lint fmt fmt-check check docker hooks build run
+
+# Makefile targets are thin shims; the implementations live in script/
+# per the scripts-to-rule-them-all pattern (see the Entrypoints section
+# of README.md). build and run are for working on the code by hand.
+
+bootstrap:
+	@script/bootstrap
+
+setup:
+	@script/setup
+
+test:
+	@script/test
+
+lint:
+	@script/lint
+
+fmt:
+	@script/fmt
+
+fmt-check:
+	@script/fmt-check
+
+check:
+	@script/check
+
+docker:
+	@script/docker
+
+hooks:
+	@script/install-precommit
+
+build:
+	@script/build
+
+run:
+	@script/run

README.md

@@ -1,18 +1,134 @@
 # smallwebwaf
 
-`smallwebwaf` is a simple, fast, logging web application firewall for people who
-host their own services. It runs inside the container of the one application it
-protects, between your reverse proxy (traefik) and the app: the app's Dockerfile
-builds `FROM` the `smallwebwaf` image, traefik sends the app's requests to
-`smallwebwaf` on port 8080, and `smallwebwaf` passes them on to the app on
-`127.0.0.1:8081`. It needs no setting, and protects the app from the first
-request with defaults chosen for a service on the open internet. It keeps its
-state in memory and in JSON files you can read and edit, and writes a detailed
-JSON log line for every request.
+`smallwebwaf` is a simple, fast, logging web application firewall, written in Go
+by [@sneak](https://sneak.berlin), for people who host their own services. It
+runs inside the container of the one application it protects, between your
+reverse proxy (traefik) and the app: the app's Dockerfile builds `FROM` the
+`smallwebwaf` image, traefik sends the app's requests to `smallwebwaf` on port
+8080, and `smallwebwaf` passes them on to the app on `127.0.0.1:8081`. It needs
+no setting, and protects the app from the first request with defaults chosen for
+a service on the open internet. It keeps its state in memory and in JSON files
+you can read and edit, and writes a detailed JSON log line for every request.
 
-Status: design stage. This repository currently holds the documents only; no
-code has been written. The design is in [`SPEC.md`](SPEC.md), and the survey of
-existing tools that led to it is in [`EVALUATION.md`](EVALUATION.md).
+Status: the first milestone is built
+(https://git.eeqj.de/sneak/smallwebwaf/issues/13). `smallwebwaf` passes each
+request to the app and the app's answer back, unchanged, within its timeouts and
+size limits, works out each client's address, and writes a JSON log line for
+every request. Rate limits per client, the country lists and the image an app
+builds on come with milestone 2
+(https://git.eeqj.de/sneak/smallwebwaf/issues/14), and the rest of the design
+after that, in the order of the build order in [`SPEC.md`](SPEC.md). The survey
+of existing tools that led to the design is in [`EVALUATION.md`](EVALUATION.md).
+
+## Getting started
+
+`smallwebwaf` is one Go binary. Until milestone 2 brings its image, build and
+run it from a clone, with Go installed:
+
+```sh
+git clone https://git.eeqj.de/sneak/smallwebwaf.git
+cd smallwebwaf
+make build
+SWWAF_UPSTREAM_URL=http://127.0.0.1:3000 ./bin/smallwebwaf
+```
+
+It then listens on port 8080 and passes every request to the app at
+`SWWAF_UPSTREAM_URL`, here an app on port 3000; with no setting at all, to an
+app on `127.0.0.1:8081`. On `SIGTERM` or `SIGINT` it stops taking requests and
+gives those in progress five seconds to finish.
+
+## What milestone 1 does
+
+- Passes each request to the app and the app's answer back unchanged: method,
+  path, query, headers, body and status. Bodies stream through in both
+  directions and are never held whole in memory. A WebSocket, or any other
+  upgraded connection, passes through, and the timeouts do not cut it.
+- Works out the client's address. A TCP peer outside `SWWAF_TRUSTED_PROXIES` is
+  the client, and the forwarded headers it sends are replaced, not passed on.
+  For a peer inside it, `X-Forwarded-For` is read from the right, and the first
+  address outside `SWWAF_TRUSTED_PROXIES` is the client; if every address in it
+  is inside, the leftmost is, and with no header the peer is. The app sees what
+  it would see from traefik directly: the same `Host`, the same
+  `X-Forwarded-Proto`, and `X-Forwarded-For` with the peer added at the end.
+- Enforces the four timeouts and the two size limits below. A limit passed
+  before the response has started gets `smallwebwaf`'s own answer: `408` for a
+  client too slow to send its request, `413` for a request body that is too
+  large, `504` for an app too slow to answer, and `502` for a response that is
+  too large or an app that cannot be reached. A request that announces a body
+  over the limit is refused before anything reaches the app. While a request
+  body is still on its way, a request timeout that runs out answers `408` if
+  `smallwebwaf` was waiting for the client to send more, and `504` if it was
+  waiting for the app to take what it had. Once the response has started, a
+  limit can only cut the connection.
+- Writes a line in the request log for each request (see "Request log" below).
+
+## Settings
+
+Each setting is an environment variable, and each has a default, so none has to
+be set. A setting that is set but invalid stops the start with a message naming
+it, and the effective settings are logged at start.
+
+- `SWWAF_LISTEN_ADDR` (default `:8080`): where `smallwebwaf` listens.
+- `SWWAF_UPSTREAM_URL` (default `http://127.0.0.1:8081`): the app, as `http` or
+  `https`, a host and an optional port, and nothing more.
+- `SWWAF_TRUSTED_PROXIES` (default `10.0.0.0/8,172.16.0.0/12,192.168.0.0/16`,
+  the private address ranges): the netblocks whose `X-Forwarded-For` is
+  believed. A list given replaces the default; set but empty, it trusts nothing.
+- `SWWAF_CLIENT_REQUEST_TIMEOUT` (default `60s`): how long a client may take to
+  send its request line and headers, and then, from the end of the headers, its
+  body.
+- `SWWAF_CLIENT_RESPONSE_TIMEOUT` (default `30m`): how long the response may
+  take to reach the client, from the end of the request to the last byte.
+- `SWWAF_UPSTREAM_REQUEST_TIMEOUT` (default `60s`): how long connecting to the
+  app and sending it the whole request may take.
+- `SWWAF_UPSTREAM_RESPONSE_TIMEOUT` (default `30m`): how long the app may take
+  to send its whole answer, from the end of the request to the last byte.
+- `SWWAF_REQUEST_MAX_BYTES` (default `100M`): the largest request body.
+- `SWWAF_RESPONSE_MAX_BYTES` (default `5G`): the largest response body.
+
+Durations are in Go's syntax, with `d` for days (`90s`, `15m`, `7d`). Sizes are
+bytes, with an optional `K`, `M` or `G`, which are powers of 1024 (`1K` is 1024
+bytes). Netblocks are in CIDR form, and a bare address stands for itself alone.
+`off` switches a timeout or a size limit off.
+
+Two limits are fixed rather than settings: the request line and headers may take
+up to 32 KiB, above which the answer is `431` and nothing reaches the app, and a
+kept-open connection that sends nothing for 120 seconds is closed. That is
+longer than the 90 seconds after which traefik closes a connection it is not
+using, so traefik never sends a request on a connection `smallwebwaf` is
+closing.
+
+## Request log
+
+`smallwebwaf` writes one JSON object per line on stdout for every request,
+refused ones included:
+
+```
+{"type":"request","time":"2026-10-03T12:00:00.123Z","client_ip":"203.0.113.9","peer_ip":"172.18.0.2","method":"GET","host":"app.example","path":"/","query":"","protocol":"HTTP/1.1","status":200,"upstream_status":200,"request_bytes":0,"response_bytes":5120,"referer":"","user_agent":"curl/8.9.1","action":"forward","duration_total":3.217,"duration_upstream_total":3.104}
+```
+
+- `time` is when the request arrived, in UTC. `peer_ip` is the TCP peer,
+  normally traefik. `path` and `query` are as the client sent them.
+- `status` is what the client was sent, `0` if nothing was; `upstream_status` is
+  what the app answered, and is left out when the app did not answer.
+- `request_bytes` and `response_bytes` count body bytes.
+- `action` is `forward` for a request passed to the app, `too_large` for a
+  request or response over its size limit, `timed_out` for one that ran out of
+  time, and `upstream_error` when the app could not be reached or its answer
+  broke off.
+- `aborted` is there, and true, when the client went away early.
+- `duration_total` and `duration_upstream_total` are in milliseconds.
+
+No body and no other header is logged. `smallwebwaf`'s own messages (start, the
+settings, stop, errors) share the stream as JSON lines marked
+`"type":"process"`.
+
+Go's HTTP server, on which `smallwebwaf` is built, reads a request's line and
+headers before `smallwebwaf` sees the request, and some requests end there,
+without a line in the log: headers over 32 KiB, which it answers `431`, headers
+slower than `SWWAF_CLIENT_REQUEST_TIMEOUT`, whose connection it closes without
+an answer, and requests it cannot read at all, which it answers itself, mostly
+with `400`.
 
 ## Why
 
@@ -249,8 +365,67 @@ visitor on your local network, another container or your monitoring, has no
 country: `SWWAF_EXCLUSIVELY_ALLOWED_COUNTRIES` refuses it unless you list it in
 `SWWAF_ALLOW_NETS`. Such addresses are never sent to GeoJS.
 
+## How the code is laid out
+
+- `cmd/smallwebwaf`: the binary, which only calls `internal/smallwebwaf`.
+- `internal/smallwebwaf`: the process: it reads the settings, listens, serves
+  requests until `SIGTERM` or `SIGINT`, and stops.
+- `internal/config`: reads the settings, the one place they are read.
+- `internal/proxy`: what happens to each request: it works out the client, runs
+  the checks, passes the request to the app and the answer back with the
+  standard library's `httputil.ReverseProxy` within the timeouts and size
+  limits, and writes the request's log line. Its `check` method is where
+  milestone 2's rate limits and country lists refuse a request.
+- `internal/requestlog`: the lines on stdout: the request log line and the
+  process's own messages.
+
+Only the Go standard library is used.
+
+## Entrypoints
+
+This repository adheres to the
+[Scripts to Rule Them All](https://github.com/github/scripts-to-rule-them-all)
+standard: the scripts in `script/` are the entrypoints for working on it, and
+the `Makefile` targets are thin shims that call them. The scripts are POSIX sh,
+so that they run in minimal containers.
+
+- `script/bootstrap`: installs what the other scripts need on the host: `make`,
+  `git`, Go for `gofmt`, node and yarn, and prettier.
+- `script/setup`: readies a fresh clone: runs `script/bootstrap`, then
+  `script/install-precommit`.
+- `script/projectname`: prints the project's name, `smallwebwaf`, which
+  `script/docker` and the others tag their images with.
+- `script/test`: runs the tests, as the `test` phase of the `Dockerfile`.
+- `script/lint`: runs golangci-lint, as the `lint` phase of the `Dockerfile`.
+- `script/fmt`: formats the Go code with `gofmt` and the Markdown with prettier.
+- `script/fmt-check`: checks the formatting, and changes nothing.
+- `script/check`: runs `script/test`, `script/lint` and `script/fmt-check`.
+- `script/docker`: builds the image, whose build runs the tests and the linter
+  first.
+- `script/cibuild`: what CI runs: `script/bootstrap`, `script/check`, then the
+  image build.
+- `script/precommit`: run by the git pre-commit hook; runs `script/check`.
+- `script/install-precommit`: installs that hook; `make hooks` runs it.
+- `script/build`: builds `bin/smallwebwaf` on the host, with Go installed, for
+  working on the code by hand; `make build` runs it.
+- `script/run`: builds `bin/smallwebwaf` with `script/build` and runs it;
+  `make run` runs it.
+
+## TODO
+
+- Milestone 2: rate limits per client, the country lists and the image an app
+  builds on (https://git.eeqj.de/sneak/smallwebwaf/issues/14).
+- The licence (https://git.eeqj.de/sneak/smallwebwaf/issues/15).
+- The rest of the design, in the order of the build order in
+  [`SPEC.md`](SPEC.md).
+
 ## Documents
 
 - [`SPEC.md`](SPEC.md): the design.
 - [`EVALUATION.md`](EVALUATION.md): what already exists, what each tool covers
   and misses, and why none was adopted.
+- [`REPO_POLICIES.md`](REPO_POLICIES.md): the policies this repository follows.
+
+## Author
+
+[@sneak](https://sneak.berlin)

REPO_POLICIES.md

@@ -0,0 +1,603 @@
+---
+title: Repository Policies
+last_modified: 2026-09-08
+---
+
+This document covers repository structure, tooling, and workflow standards. Code
+style conventions are in separate documents:
+
+- [Code Styleguide](https://git.eeqj.de/sneak/prompts/raw/branch/main/prompts/CODE_STYLEGUIDE.md)
+  (general, bash, Docker)
+- [Go](https://git.eeqj.de/sneak/prompts/raw/branch/main/prompts/CODE_STYLEGUIDE_GO.md)
+- [JavaScript](https://git.eeqj.de/sneak/prompts/raw/branch/main/prompts/CODE_STYLEGUIDE_JS.md)
+- [Python](https://git.eeqj.de/sneak/prompts/raw/branch/main/prompts/CODE_STYLEGUIDE_PYTHON.md)
+- [Go HTTP Server Conventions](https://git.eeqj.de/sneak/prompts/raw/branch/main/prompts/GO_HTTP_SERVER_CONVENTIONS.md)
+
+---
+
+- Cross-project documentation (such as this file) must include
+  `last_modified: YYYY-MM-DD` in the YAML front matter so it can be kept in sync
+  with the authoritative source as policies evolve.
+
+- **ALL external references must be pinned by cryptographic hash.** This
+  includes Docker base images, Go modules, npm packages, GitHub Actions, and
+  anything else fetched from a remote source. Version tags (`@v4`, `@latest`,
+  `:3.21`, etc.) are server-mutable and therefore remote code execution
+  vulnerabilities. The ONLY acceptable way to reference an external dependency
+  is by its content hash (Docker `@sha256:...`, Go module hash in `go.sum`, npm
+  integrity hash in lockfile, GitHub Actions `@<commit-sha>`). No exceptions.
+  This also means never `curl | bash` to install tools like pyenv, nvm, rustup,
+  etc. Instead, download a specific release archive from GitHub, verify its hash
+  (hardcoded in the Dockerfile or script), and only then install. Unverified
+  install scripts are arbitrary remote code execution. This is the single most
+  important rule in this document. Double-check every external reference in
+  every file before committing. There are zero exceptions to this rule.
+
+- Every repo with software must have a root `Makefile` with these targets:
+  `make bootstrap`, `make setup`, `make test`, `make lint`, `make fmt` (writes),
+  `make fmt-check` (read-only), `make check` (runs `test`, `lint`, `fmt-check`),
+  `make docker`, and `make hooks` (installs pre-commit hook). A model Makefile
+  is at `https://git.eeqj.de/sneak/prompts/raw/branch/main/Makefile`.
+
+- Repos follow the
+  [Scripts to Rule Them All](https://github.com/github/scripts-to-rule-them-all)
+  pattern: the implementation of each Makefile target lives in an executable
+  script in `script/` (`script/bootstrap`, `script/setup`, `script/test`,
+  `script/lint`, `script/fmt`, `script/fmt-check`, `script/check`,
+  `script/docker`), and the Makefile targets are thin shims that call them. The
+  scripts must be POSIX sh (`#!/bin/sh`, `set -eu`, no bashisms) so they run in
+  minimal containers (e.g. alpine images have no bash); locate the repo root
+  with `$(cd "$(dirname "$0")/.." && pwd -P)` and `cd` there before acting. From
+  the standard's canonical set we use `bootstrap`, `setup` (make the repo ready
+  for development after a fresh clone: runs `bootstrap`, then
+  `install-precommit`, plus any repo-specific initialization), `test`, and
+  `cibuild`. `script/bootstrap` installs all dependencies idempotently and
+  assumes nothing is present: base tools come from nix, apt, brew, or apk
+  (detected in that order; apt runs noninteractive). For node it uses the
+  installed node if present; otherwise it installs a PINNED node version via
+  nvm, first installing nvm itself if missing — from a hash-verified GitHub
+  release archive (never `curl | sh`), with bash installed as an explicit
+  prerequisite since nvm requires bash. yarn is then pinned via
+  `corepack prepare yarn@<version> --activate`. Never install "latest" or "lts";
+  always exact versions. `script/cibuild` runs the CI build: it changes to the
+  repo root, runs `script/bootstrap`, runs `script/check`, and builds the image
+  with the version; the Gitea workflow calls it. **`script/cibuild` runs
+  `script/bootstrap` first**, because the workflow checks out the repo and runs
+  nothing else, while `script/fmt-check` runs the formatter on the host: on a
+  pristine checkout with nothing installed the run dies there, after the
+  containerised gates have passed. **The bootstrap alone is not enough**:
+  `script/bootstrap` installs node and yarn under nvm and leaves neither on the
+  `PATH` of the shell that called it, so a bare `yarn` still exits 127. The host
+  entrypoints that need yarn — `script/fmt` and `script/fmt-check` — therefore
+  source nvm for the pinned node version before invoking it, exactly as
+  `script/bootstrap`'s own install step does. A runner carrying nothing but
+  docker and git then gets through `script/check`. Four further scripts are our
+  own extensions to the standard: `script/check` runs `script/test`,
+  `script/lint` and `script/fmt-check`; `script/precommit` is what the git
+  pre-commit hook runs, and it calls `script/check`; `script/install-precommit`
+  installs the git pre-commit hook (the `make hooks` target shims to it); and
+  `script/projectname` (literally that filename) simply outputs the project's
+  name. Scripts that need the name call `script/projectname` — e.g.
+  `script/docker` assembles its image tag from it — so those scripts stay
+  byte-identical across all repos. Repo-type-specific pre-commit extras (e.g.
+  `go mod tidy` verification in Go repos) belong in `script/precommit`, not in
+  the hook itself. Model scripts are at
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/script/<name>`. The README
+  must document the provided scripts in an **Entrypoints** section (see the
+  README requirements below).
+
+- Always use Makefile targets (`make fmt`, `make test`, `make lint`, etc.)
+  instead of invoking the underlying tools directly. The Makefile is the single
+  source of truth for how these operations are run.
+
+- The Makefile is authoritative documentation for how the repo is used. Beyond
+  the required targets above, it should have targets for every common operation:
+  running a local development server (`make run`, `make dev`), re-initializing
+  or migrating the database (`make db-reset`, `make migrate`), building
+  artifacts (`make build`), generating code, seeding data, or anything else a
+  developer would do regularly. If someone checks out the repo and types
+  `make<tab>`, they should see every meaningful operation available. A new
+  contributor should be able to understand the entire development workflow by
+  reading the Makefile.
+
+- Every repo should have a `Dockerfile`, and it carries the repo's gates: a
+  `lint` phase and a `test` phase, with the final stage depending on both so the
+  image cannot be built unless they pass. For non-server repos the final stage
+  brings up a development environment; for server repos it is the runtime image.
+  Dockerfiles install development prerequisites by running `script/bootstrap`
+  rather than duplicating installs inline; COPY `script/` and the dependency
+  manifests (`package.json` + `yarn.lock`, `go.mod` + `go.sum`, etc.) before
+  running it.
+
+- **Linting and testing run in Docker, as phases of the `Dockerfile`.** There is
+  no separate lint file. `script/lint` and `script/test` each build one phase
+  and nothing else:
+
+    ```sh
+    docker build --no-cache --target lint -t "$(script/projectname)-lint" .
+    docker build --no-cache --target test -t "$(script/projectname)-test" .
+    ```
+
+    **A stage that is not the last one in the file is built only when the final
+    stage's chain depends on it, or when `--target` names it.** That is why the
+    two gates are always invoked by name here, and why the final stage carries a
+    `COPY --from=` of a harmless file from each of them: without that edge a
+    plain `docker build .` builds the last stage alone and exits 0 having linted
+    and tested nothing.
+
+    **Every `docker build` in `script/` is tagged**, here and in
+    `script/cibuild` and `script/docker`. An untagged build leaves a dangling
+    image behind on every invocation, on every developer host and every CI
+    runner; a tagged one replaces the previous image.
+
+    Inside a phase the tool is invoked directly — `golangci-lint`, `go test`,
+    `eslint`, `prettier` — never through `make lint` or `script/test`, which are
+    themselves a `docker build` and would recurse into a daemon that does not
+    exist in a build step. Formatting is the exception and stays on the host:
+    `script/fmt` writes the working tree, and `script/fmt-check` is its
+    read-only twin.
+
+    **No lint verdict may come from a host invocation of the linter.** On a
+    shared host golangci-lint reads a result cache keyed on file content rather
+    than location, so a second checkout of the same content is served the first
+    one's findings, and a host-global lock in `$TMPDIR` makes concurrent runs
+    exit non-zero with `parallel golangci-lint is running` — a status a caller
+    cannot tell from real findings. Both have produced wrong verdicts in this
+    org, in both directions. A container has its own cache, its own `TMPDIR` and
+    a digest-pinned binary, so neither is reachable.
+
+- **Any build that runs checks is built with `--no-cache`.** Docker invalidates
+  a `COPY` layer only when the copied content changes, so on an unchanged tree
+  the check `RUN` is served from cache, nothing executes, and the build still
+  exits 0. Every `docker build` in `script/` therefore passes `--no-cache`:
+  `script/lint`, `script/test`, `script/cibuild` and `script/docker` are the
+  four, and there is no fifth — `script/check` runs the two gate phases and
+  `script/fmt-check`, and builds no image of its own. A bare `docker build .` is
+  not evidence that anything ran: a sub-second build reporting success is a
+  cache hit, not a result. Never invalidate by pruning — `docker builder prune`
+  and friends destroy a build cache shared with every other build on the host.
+
+- **The gate phases are separate stages, and the build stage depends on both.**
+  The lint phase is based on the `golangci/golangci-lint` image (pinned by
+  hash), so lint failures surface in seconds rather than after a full compile,
+  and the test phase is based on the Go image. The canonical Go repo
+  `Dockerfile`:
+
+    ```dockerfile
+    # Lint phase
+    # golangci/golangci-lint:v2.x.x, YYYY-MM-DD
+    FROM golangci/golangci-lint@sha256:... AS lint
+    WORKDIR /src
+    COPY go.mod go.sum ./
+    RUN go mod download
+    COPY . .
+    RUN golangci-lint run --config .golangci.yml ./...
+
+    # Test phase
+    # golang:1.x-alpine, YYYY-MM-DD
+    FROM golang@sha256:... AS test
+    WORKDIR /src
+    COPY go.mod go.sum ./
+    RUN go mod download
+    COPY . .
+    RUN go test -timeout 90s -race -cover ./... || \
+        { echo "--- Rerunning with -v for details ---"; \
+          go test -timeout 90s -race -v ./...; exit 1; }
+
+    # Build stage. Nothing is wanted from either phase above; the copies
+    # are what make BuildKit build them first, so this stage cannot run
+    # unless lint and test passed.
+    # golang:1.x-alpine, YYYY-MM-DD
+    FROM golang@sha256:... AS builder
+    COPY --from=lint /src/go.sum /dev/null
+    COPY --from=test /src/go.sum /dev/null
+    WORKDIR /src
+    COPY go.mod go.sum ./
+    RUN go mod download
+    COPY . .
+
+    ARG VERSION=dev
+    RUN CGO_ENABLED=0 go build -trimpath \
+        -ldflags="-s -w -X main.Version=${VERSION}" \
+        -o /app ./cmd/app/
+
+    # Runtime stage, and the last one
+    FROM alpine@sha256:...
+    COPY --from=builder /app /usr/local/bin/app
+    ENTRYPOINT ["app"]
+    ```
+
+    Key points:
+    - The lint phase uses the `golangci/golangci-lint` image directly (it has
+      both Go and the linter), so nothing needs installing.
+    - `COPY --from=<phase> /src/go.sum /dev/null` is a no-op copy whose only
+      purpose is the ordering edge. BuildKit runs stages in parallel by default,
+      and a stage nothing depends on is not built at all, so without these two
+      lines a red gate would not fail the build.
+    - Keep the runtime stage last, and if you add a stage after it, give it the
+      same two copies. A plain `docker build .` builds the last stage's chain
+      and nothing else.
+    - If the project uses `//go:embed` directives that reference build artifacts
+      (e.g. a web frontend compiled in a separate stage), the lint phase must
+      create placeholder files so the embed directives resolve. Example:
+      `RUN mkdir -p web/dist && touch web/dist/index.html web/dist/style.css`.
+    - If the project requires CGO or system libraries for linting (e.g.
+      `vips-dev`), install them in the lint phase with `apk add`.
+    - `ARG VERSION=dev` is declared in the stage that compiles and supplied by
+      `script/docker` and `script/cibuild`; no stage may call `git describe`.
+
+- Every repo should have a Gitea Actions workflow (`.gitea/workflows/`) that
+  runs `script/cibuild` on push, and checks out the repo as its only other step.
+  That script bootstraps, runs the gate phases, and then builds the image, so a
+  successful run means every check passed; a bare `docker build .` does not
+  carry the same guarantee, because its gate phases may come from the cache. The
+  image build is uncached and so runs the gate phases a second time. That is the
+  price of the rule above, and it is worth paying: the image that ships is built
+  from a run of its own gates rather than from a cache entry.
+
+- Use platform-standard formatters: `black` for Python, `prettier` for
+  JS/CSS/Markdown/HTML, `go fmt` for Go. Always use default configuration with
+  two exceptions: four-space indents (except Go), and `proseWrap: always` for
+  Markdown (hard-wrap at 80 columns). Documentation and writing repos (Markdown,
+  HTML, CSS) should also have `.prettierrc` and `.prettierignore`.
+
+- Pre-commit hook: runs `script/precommit`, which calls `script/check`. If local
+  testing is not possible in the repo, `script/precommit` may skip `script/test`
+  and run only `script/lint` and `script/fmt-check`. The hook is installed by
+  `script/install-precommit`; the Makefile must provide a `make hooks` target
+  that shims to it.
+
+- All repos with software must have tests that run via the platform-standard
+  test framework (`go test`, `pytest`, `jest`/`vitest`, etc.). If no meaningful
+  tests exist yet, add the most minimal test possible — e.g. importing the
+  module under test to verify it compiles/parses. There is no excuse for
+  `make test` to be a no-op.
+
+- `make test` must complete in under 60 seconds. That is the hard cap, and a
+  suite that exceeds it fails. Under 20 seconds is the target. A suite between
+  20 and 60 seconds is still green, but the overage must be filed as an
+  improvement bug against that repo. Add a 90-second timeout to the test
+  invocation (`go test -timeout 90s`). The backstop deliberately sits above the
+  hard cap so that it catches a genuinely hung test rather than a merely slow
+  one.
+
+- **The test command should use the conditional verbose rerun pattern.** Run
+  tests without `-v` (verbose) first. If tests fail, automatically rerun with
+  `-v` to show full output. This keeps CI logs and `docker build` output clean
+  on success (just package/suite summaries) while providing full diagnostic
+  detail on failure (every test case, every assertion). The command lives in the
+  `test` phase of the `Dockerfile`, since `script/test` builds that phase; the
+  Makefile form below is the same pattern for any repo-local invocation:
+
+    ```makefile
+    test:
+    	@<test-command> || \
+    		{ echo "--- Rerunning with -v for details ---"; \
+    		  <test-command-with-v>; exit 1; }
+    ```
+
+    Go example:
+
+    ```makefile
+    test:
+    	@go test -count=1 -timeout 90s -race -cover ./... || \
+    		{ echo "--- Rerunning with -v for details ---"; \
+    		  go test -count=1 -timeout 90s -race -v ./...; exit 1; }
+    ```
+
+    `-count=1` is required on both invocations: it defeats Go's test _result_
+    cache, so the target cannot report a pass it did not earn, and the rerun
+    reproduces a failure instead of replaying it. It leaves the build cache
+    alone, so it costs the runtime of the suite and no recompilation.
+
+    Note that this is a second, independent cache, stacked below the Docker
+    layer cache that [issue #26](https://git.eeqj.de/sneak/prompts/issues/26)
+    addresses. `CHECK_EPOCH` guarantees the `RUN make test` _step_ re-executes;
+    it does not guarantee `go test` inside that step does any work, because the
+    `GOCACHE` baked into earlier image layers survives into the re-executed
+    step. They are two separate defects requiring two separate fixes, and a fix
+    for one must not be recorded as covering the other.
+
+    Python example:
+
+    ```makefile
+    test:
+    	@python -m pytest || \
+    		{ echo "--- Rerunning with -v for details ---"; \
+    		  python -m pytest -v; exit 1; }
+    ```
+
+    The `exit 1` ensures the target always fails after a rerun — the first run
+    already proved the tests are broken, so the build must not pass even if a
+    flaky test happens to succeed on the second attempt. The rerun exists solely
+    for diagnostic output.
+
+- Docker builds must complete in under 5 minutes.
+
+- `make check` must not modify any files in the repo. Tests may use temporary
+  directories.
+
+- `main` must always pass `make check`, no exceptions.
+
+- Never commit secrets. `.env` files, credentials, API keys, and private keys
+  must be in `.gitignore`. No exceptions.
+
+- `.gitignore` should be comprehensive from the start: OS files (`.DS_Store`),
+  editor files (`.swp`, `*~`), in-repo agent scratch directories (`.claude/`),
+  language build artifacts, and `node_modules/`. Fetch the standard `.gitignore`
+  from `https://git.eeqj.de/sneak/prompts/raw/branch/main/.gitignore` when
+  setting up a new repo. These patterns are written to `.gitignore`'s own
+  semantics, in which an unanchored pattern already matches at every depth; they
+  are not a `.dockerignore` and must not be transplanted into one unmodified.
+
+- **`.dockerignore` does not use `.gitignore` semantics, and copying patterns
+  across unmodified leaves secrets in the build context.** Docker matches with
+  `moby/patternmatcher`: `filepath.Match` semantics plus a `**` extension, so
+  `*` does not cross `/` and a pattern without a leading `**/` is anchored at
+  the build-context root. A `.dockerignore` listing `.env`, `*.pem` and `*.key`
+  therefore excludes only the copies at the repository root, while `config/.env`
+  and `certs/server.key` still reach the context and can land in an image layer
+  — which is more dangerous than a short file with no secret patterns at all,
+  because it reads as solved and stops anyone looking. Give every
+  depth-independent pattern the `**/` prefix and leave only genuinely
+  root-anchored entries unprefixed: `.git`, and the repo's own host-built
+  binary, written `/myapp` and never `**/myapp`, which would also match
+  `cmd/myapp/` and delete the package directory from the context. Matching is
+  case-sensitive, and an ALL-CAPS twin per pattern still misses `Server.Key`, so
+  secret names use character ranges — `**/*.[kK][eE][yY]`, `**/*.[pP][eE][mM]`,
+  and likewise for `.envrc` and the extensionless SSH keys. Where such a pattern
+  also catches something the build needs, re-include it with a negation
+  (`!docs/example.env`); deleting the pattern reopens the exposure for every
+  other file it covers. Fetch the standard `.dockerignore` from
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.dockerignore` and extend
+  it with the repo's own artifacts.
+
+- **In-repo agent scratch belongs in both files, written to each file's own
+  semantics.** `.claude/` holds one worktree per in-flight agent — an entire
+  additional checkout of the repo — so under `COPY . .` the build context
+  inflates by a multiple of the repo and another session's unreviewed work can
+  be copied into an image layer. In `.gitignore` the entry is `.claude/`,
+  unanchored. In `.dockerignore` it is `.claude`, anchored and with **no** `**/`
+  prefix, because the prefixed form would also delete any nested directory of
+  that name from the build. Anchoring carries a known gap that the canonical
+  `.dockerignore` states in its own comment, since consuming repos receive the
+  file and not the tracker: the directory is created in the agent's working
+  directory, so a repo running agents in subdirectories still ships
+  `services/api/.claude/` and must add its own anchored entry there.
+
+- **Excluding `.git` means `git describe` cannot run inside any build stage, and
+  it fails quietly there.** In a build stage there is no repository, so
+  `git describe` writes nothing to stdout, `-X main.Version=` comes out empty,
+  the binary reports no version at all, and the build still exits 0. Compute the
+  version on the host and thread it in as a build arg. `script/docker` and
+  `script/cibuild` do this, byte-identically across repos:
+
+    ```sh
+    # Own line: a failing command substitution inside an argument does not
+    # trip `set -e`, so the inline form degrades to an empty constant.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$(script/projectname)" .
+    ```
+
+    `--always` makes an untagged repo yield an abbreviated commit hash rather
+    than failing, and the `[ -n "$version" ]` line is the single place the
+    fallback is applied — a live check that fires on a build from an export with
+    no `.git` and on a repository with no commits yet. Do not fold it into the
+    substitution as `|| echo unknown`, which makes the guard unreachable. The
+    Dockerfile's side is `ARG VERSION=dev` in the stage that compiles, declared
+    there because `ARG` is stage-scoped; passing `VERSION` to a repo whose
+    Dockerfile declares no such `ARG` is ignored and costs nothing, which is why
+    the scripts stay byte-identical. One consequence for CI: the standard
+    checkout action clones shallow and fetches no tags, so a repo that embeds a
+    tag-derived version must set `fetch-depth: 0` on its checkout step.
+
+- **Verify `.dockerignore` by enumerating the image, not by reading the
+  patterns.** Plant files at the root _and_ at least two directories deep, build
+  a probe image that does `COPY . .`, and list what actually landed
+  (`docker run --rm --entrypoint find IMAGE /app`). The `transferring context`
+  size is not a substitute: a nested secret is a few bytes, and BuildKit
+  transfers only the delta from the previous build.
+
+- **No build artifacts in version control.** Code-derived data (compiled
+  bundles, minified output, generated assets) must never be committed to the
+  repository if it can be avoided. The build process (e.g. Dockerfile, Makefile)
+  should generate these at build time. Notable exception: Go protobuf generated
+  files (`.pb.go`) ARE committed because repos need to work with `go get`, which
+  downloads code but does not execute code generation.
+
+- Never use `git add -A` or `git add .`. Always stage files explicitly by name.
+
+- Never force-push to `main`.
+
+- Make all changes on a feature branch. You can do whatever you want on a
+  feature branch.
+
+- `.golangci.yml` is standardized. The vendored copy in a consuming repo must
+  _NEVER_ be modified by an agent: fetch it from
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.golangci.yml` and keep it
+  byte-identical, so that no repo can quietly loosen its own linting. Linter
+  configuration changes are made to the canonical copy in the `prompts` repo and
+  reach consuming repos by re-vendoring; an agent may open a PR against
+  canonical, which only the user merges. One list is exempt from byte-identity,
+  because it cannot be written once for every repo: the `deny` list of the
+  `test-support` depguard rule, where a repo names its own test-support packages
+  by full import path. A repo adds entries there and changes nothing else, and a
+  re-vendor carries its entries forward. The canonical golangci-lint version is
+  v2.12.2 (released 2026-05-06), pinned as the digest of the lint phase's base
+  image
+  (`golangci/golangci-lint@sha256:5cceeef04e53efe1470638d4b4b4f5ceefd574955ab3941b2d9a68a8c9ad5240`,
+  which reports `2.12.2 built with go1.26.2 from c0d3ddc9`). That digest is the
+  only pin, since no repo installs golangci-lint on the host: bumping the
+  version means changing it and nothing else.
+
+- **`script/bootstrap` installs a pinned tool by comparing versions, never by
+  testing presence.** An `if ! command -v <tool>; then install; fi` guard tests
+  `PATH` only, so on an already-provisioned machine the pin is inert and a
+  version bump is a silent no-op — while the Dockerfile, installing into a clean
+  image, gets the pinned version, so a local `make check` and `make docker` can
+  disagree about what the tool even is. The canonical form:
+    - compares the installed version against the pin over the **whole** version
+      token; a parser that stops at the first `-` reports `2.12.2` for a host
+      running `2.12.2-rc1` and skips the install;
+    - treats absent, non-zero, empty or unrecognised `--version` output as a
+      mismatch, so the failure direction is a redundant install and never a
+      skipped one;
+    - after installing, re-resolves the binary the way callers do — `hash -r`,
+      then through `PATH`, not through the directory the installer wrote to —
+      and fails naming the resolved path, since an install that a shadowing
+      binary hides succeeds while changing nothing any caller sees;
+    - is actually called, and prints the version on both success paths: a
+      function defined and never invoked has the same exit status and the same
+      empty output as one that worked.
+
+    Keep it POSIX sh: no arrays, no `[[`, no `grep -P`.
+
+- When pinning images or packages by hash, add a comment above the reference
+  with the version and date (YYYY-MM-DD).
+
+- Use `yarn`, not `npm`.
+
+- Write all dates as YYYY-MM-DD (ISO 8601).
+
+- Simple projects should be configured with environment variables.
+
+- Dockerized web services listen on port 8080 by default, overridable with
+  `PORT`.
+
+- **HTTP/web services must be hardened for production internet exposure before
+  tagging 1.0.** This means full compliance with security best practices
+  including, without limitation, all of the following:
+    - **Security headers** on every response:
+        - `Strict-Transport-Security` (HSTS) with `max-age` of at least one year
+          and `includeSubDomains`.
+        - `Content-Security-Policy` (CSP) with a restrictive default policy
+          (`default-src 'self'` as a baseline, tightened per-resource as
+          needed). Never use `unsafe-inline` or `unsafe-eval` unless
+          unavoidable, and document the reason.
+        - `X-Frame-Options: DENY` (or `SAMEORIGIN` if framing is required).
+          Prefer the `frame-ancestors` CSP directive as the primary control.
+        - `X-Content-Type-Options: nosniff`.
+        - `Referrer-Policy: strict-origin-when-cross-origin` (or stricter).
+        - `Permissions-Policy` restricting access to browser features the
+          application does not use (camera, microphone, geolocation, etc.).
+    - **Request and response limits:**
+        - Maximum request body size enforced on all endpoints (e.g. Go
+          `http.MaxBytesReader`). Choose a sane default per-route; never accept
+          unbounded input.
+        - Maximum response body size where applicable (e.g. paginated APIs).
+        - `ReadTimeout` and `ReadHeaderTimeout` on the `http.Server` to defend
+          against slowloris attacks.
+        - `WriteTimeout` on the `http.Server`.
+        - `IdleTimeout` on the `http.Server`.
+        - Per-handler execution time limits via `context.WithTimeout` or
+          chi/stdlib `middleware.Timeout`.
+    - **Authentication and session security:**
+        - Rate limiting on password-based authentication endpoints. API keys are
+          high-entropy and not susceptible to brute force, so they are exempt.
+        - CSRF tokens on all state-mutating HTML forms. API endpoints
+          authenticated via `Authorization` header (Bearer token, API key) are
+          exempt because the browser does not attach these automatically.
+        - Passwords stored using bcrypt, scrypt, or argon2 — never plain-text,
+          MD5, or SHA.
+        - Session cookies set with `HttpOnly`, `Secure`, and `SameSite=Lax` (or
+          `Strict`) attributes.
+    - **Reverse proxy awareness:**
+        - True client IP detection when behind a reverse proxy
+          (`X-Forwarded-For`, `X-Real-IP`). The application must accept
+          forwarded headers only from a configured set of trusted proxy
+          addresses — never trust `X-Forwarded-For` unconditionally.
+    - **CORS:**
+        - Authenticated endpoints must restrict `Access-Control-Allow-Origin` to
+          an explicit allowlist of known origins. Wildcard (`*`) is acceptable
+          only for public, unauthenticated read-only APIs.
+    - **Error handling:**
+        - Internal errors must never leak stack traces, SQL queries, file paths,
+          or other implementation details to the client. Return generic error
+          messages in production; detailed errors only when `DEBUG` is enabled.
+    - **TLS:**
+        - Services never terminate TLS directly. They are always deployed behind
+          a TLS-terminating reverse proxy. The service itself listens on plain
+          HTTP. However, HSTS headers and `Secure` cookie flags must still be
+          set by the application so that the browser enforces HTTPS end-to-end.
+
+    This list is non-exhaustive. Apply defense-in-depth: if a standard security
+    hardening measure exists for HTTP services and is not listed here, it is
+    still expected. When in doubt, harden.
+
+- `README.md` is the primary documentation. Required sections:
+    - **Description**: First line must include the project name, purpose,
+      category (web server, SPA, CLI tool, etc.), license, and author. Example:
+      "µPaaS is an MIT-licensed Go web application by @sneak that receives
+      git-frontend webhooks and deploys applications via Docker in realtime."
+    - **Getting Started**: Copy-pasteable install/usage code block.
+    - **Entrypoints**: Opens by stating that the repo adheres to the
+      [Scripts to Rule Them All](https://github.com/github/scripts-to-rule-them-all)
+      standard (with that link), then documents each provided `script/`
+      entrypoint and its purpose.
+    - **Rationale**: Why does this exist?
+    - **Design**: How is the program structured?
+    - **TODO**: Update meticulously, even between commits. When planning, put
+      the todo list in the README so a new agent can pick up where the last one
+      left off.
+    - **License**: MIT, GPL, or WTFPL. Ask the user for new projects. Include a
+      `LICENSE` file in the repo root and a License section in the README.
+    - **Author**: [@sneak](https://sneak.berlin).
+
+- First commit of a new repo should contain only `README.md`.
+
+- Go module root: `sneak.berlin/go/<name>`. Always run `go mod tidy` before
+  committing.
+
+- Use SemVer.
+
+- Database migrations live in `internal/db/migrations/` and must be embedded in
+  the binary.
+    - `000_migration.sql` — contains ONLY the creation of the migrations
+      tracking table itself. Nothing else.
+    - `001_schema.sql` — the full application schema.
+    - **Pre-1.0.0:** never add additional migration files (002, 003, etc.).
+      There is no installed base to migrate. Edit `001_schema.sql` directly.
+    - **Post-1.0.0:** add new numbered migration files for each schema change.
+      Never edit existing migrations after release.
+
+- All repos should have an `.editorconfig` enforcing the project's indentation
+  settings.
+
+- Avoid putting files in the repo root unless necessary. Root should contain
+  only project-level config files (`README.md`, `Makefile`, `Dockerfile`,
+  `LICENSE`, `.gitignore`, `.editorconfig`, `REPO_POLICIES.md`, and
+  language-specific config). Everything else goes in a subdirectory. Canonical
+  subdirectory names:
+    - `bin/` — executable scripts and tools
+    - `cmd/` — Go command entrypoints; thin only: one `main.go` per binary whose
+      body is a single call into `internal/` or `pkg/`, no project logic in
+      `cmd/`
+    - `configs/` — configuration templates and examples
+    - `deploy/` — deployment manifests (k8s, compose, terraform)
+    - `docs/` — documentation and markdown (README.md stays in root)
+    - `internal/` — Go internal packages
+    - `internal/db/migrations/` — database migrations
+    - `pkg/` — Go library packages
+    - `share/` — systemd units, data files
+    - `static/` — static assets (images, fonts, etc.)
+    - `web/` — web frontend source
+
+- When setting up a new repo, files from the `prompts` repo may be used as
+  templates. Fetch them from
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/<path>`.
+
+- New repos must contain at minimum:
+    - `README.md`, `.git`, `.gitignore`, `.editorconfig`
+    - `LICENSE`, `REPO_POLICIES.md` (copy from the `prompts` repo)
+    - `Makefile`
+    - `script/` entrypoints (`bootstrap`, `setup`, `projectname`, `test`,
+      `lint`, `fmt`, `fmt-check`, `check`, `docker`, `cibuild`, `precommit`,
+      `install-precommit`)
+    - `Dockerfile`, `.dockerignore`
+    - `.gitea/workflows/check.yml`
+    - Go: `go.mod`, `go.sum`, `.golangci.yml`
+    - JS: `package.json`, `yarn.lock`, `.prettierrc`, `.prettierignore`
+    - Python: `pyproject.toml`

SPEC.md

@@ -1,8 +1,8 @@
 # smallwebwaf SPEC (draft): protective reverse proxy for one app
 
-Status: fourth draft, with the owner's rulings to date applied. Nothing has been
-built yet. `EVALUATION.md` beside this file explains why no existing tool was
-chosen.
+Status: fourth draft, with the owner's rulings to date applied. Milestone 1 of
+the build order is built. `EVALUATION.md` beside this file explains why no
+existing tool was chosen.
 
 ## Purpose
 
@@ -409,7 +409,8 @@ The settings, by group:
   Bodies stream straight through, so a request body reaches the app while the
   client is still sending it.
     - `SWWAF_CLIENT_REQUEST_TIMEOUT` (default `60s`): how long a client may take
-      to send its whole request, headers and body.
+      to send its request line and headers, and then, from the end of the
+      headers, its body.
     - `SWWAF_CLIENT_REQUEST_HEADER_MAX_BYTES` (default `32K`): the largest
       request line and headers a client may send. Over it, `smallwebwaf` answers
       `431` and closes the connection, and nothing reaches the app.
@@ -436,6 +437,12 @@ The settings, by group:
       an app that is too slow. A request that announces a body larger than its
       limit is refused before anything reaches the app. Once the response has
       started it can only be cut off, and the connection is closed.
+    - Go's HTTP server, on which `smallwebwaf` is built, reads a request's line
+      and headers before `smallwebwaf` sees the request. A client that takes
+      longer than `SWWAF_CLIENT_REQUEST_TIMEOUT` to send them gets no answer:
+      the server closes its connection. Headers over
+      `SWWAF_CLIENT_REQUEST_HEADER_MAX_BYTES` are answered `431` by the server
+      itself. Neither request gets a line in the request log.
     - A WebSocket connection leaves these limits behind once it is upgraded: it
       stays open until either side closes it.
 - Lookup of AS number and country (R7). On by default through GeoJS, which needs
@@ -1012,10 +1019,12 @@ and the running `smallwebwaf` takes the edit in.
 
 ## Request log
 
-One JSON object per line on stdout for every request, including refused ones.
-stdout is always on. When `SWWAF_LOG_REMOTE_URL` is set the same lines are also
-sent to the remote endpoint, so a deployment can stop depending on docker's log
-handling while `docker logs` keeps working.
+One JSON object per line on stdout for every request, including refused ones,
+apart from those Go's HTTP server ends before `smallwebwaf` sees them (see
+"Configuration surface", size and time limits). stdout is always on. When
+`SWWAF_LOG_REMOTE_URL` is set the same lines are also sent to the remote
+endpoint, so a deployment can stop depending on docker's log handling while
+`docker logs` keeps working.
 
 - Standard web log fields: `time` (RFC 3339 with milliseconds), `instance`,
   `client_ip`, `method`, `scheme`, `host`, `path`, `query`, `protocol`,

cmd/smallwebwaf/main.go

@@ -0,0 +1,18 @@
+// Command smallwebwaf is a web application firewall for one app: it runs
+// between traefik and the app, and passes requests through within its
+// limits. Everything it does is in internal/smallwebwaf.
+package main
+
+import (
+	"os"
+
+	"sneak.berlin/go/smallwebwaf/internal/smallwebwaf"
+)
+
+// Version is the version of the binary, set when it is built with
+// -ldflags "-X main.Version=...".
+var Version = "dev" //nolint:gochecknoglobals // the linker sets it
+
+func main() {
+	os.Exit(smallwebwaf.Main(Version))
+}

go.mod

@@ -0,0 +1,3 @@
+module sneak.berlin/go/smallwebwaf
+
+go 1.26.0

internal/config/config.go

@@ -0,0 +1,352 @@
+// Package config reads smallwebwaf's settings. Every setting is an
+// environment variable whose name starts with SWWAF_, every setting has a
+// default, and this package is the one place they are read.
+package config
+
+import (
+	"errors"
+	"fmt"
+	"log/slog"
+	"math"
+	"net"
+	"net/netip"
+	"net/url"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// Config is smallwebwaf's settings. A timeout or size of zero is off.
+type Config struct {
+	// ListenAddr is where smallwebwaf listens (SWWAF_LISTEN_ADDR).
+	ListenAddr string
+	// UpstreamURL is the app (SWWAF_UPSTREAM_URL).
+	UpstreamURL *url.URL
+	// TrustedProxies are the netblocks whose X-Forwarded-For is
+	// believed (SWWAF_TRUSTED_PROXIES).
+	TrustedProxies []netip.Prefix
+	// ClientRequestTimeout bounds reading the whole request from the
+	// client (SWWAF_CLIENT_REQUEST_TIMEOUT).
+	ClientRequestTimeout time.Duration
+	// ClientResponseTimeout bounds writing the whole response to the
+	// client (SWWAF_CLIENT_RESPONSE_TIMEOUT).
+	ClientResponseTimeout time.Duration
+	// UpstreamRequestTimeout bounds connecting to the app and writing
+	// the whole request to it (SWWAF_UPSTREAM_REQUEST_TIMEOUT).
+	UpstreamRequestTimeout time.Duration
+	// UpstreamResponseTimeout bounds reading the whole response from
+	// the app (SWWAF_UPSTREAM_RESPONSE_TIMEOUT).
+	UpstreamResponseTimeout time.Duration
+	// RequestMaxBytes is the largest request body
+	// (SWWAF_REQUEST_MAX_BYTES).
+	RequestMaxBytes int64
+	// ResponseMaxBytes is the largest response body
+	// (SWWAF_RESPONSE_MAX_BYTES).
+	ResponseMaxBytes int64
+
+	// settings are the values read, as given or by default, for the
+	// log line at start.
+	settings []slog.Attr
+}
+
+// off is the value that switches a timeout or a size limit off.
+const off = "off"
+
+const (
+	day      = 24 * time.Hour
+	kibibyte = 1 << 10
+	mebibyte = 1 << 20
+	gibibyte = 1 << 30
+)
+
+var (
+	errNotDuration = errors.New(
+		"is not a duration such as 90s, 15m or 7d, or off")
+	errNotSize = errors.New(
+		"is not a size such as 512K, 100M or 5G, or off")
+	errNotPositive = errors.New("must be more than zero, or off")
+	errEmptyItem   = errors.New("has an empty item in its list")
+	errNotNetblock = errors.New(
+		"is not a netblock such as 10.0.0.0/8, or an address")
+	errNotListenAddr = errors.New(
+		"is not an address to listen on, such as :8080")
+	errNotUpstreamURL = errors.New(
+		"is not a URL with only a scheme, a host and an optional port, " +
+			"such as http://127.0.0.1:8081")
+)
+
+// FromEnvironment reads the settings with lookupEnv, normally
+// os.LookupEnv. A setting that is not set takes its default. A setting
+// that is set but invalid is an error that names it.
+func FromEnvironment(lookupEnv func(string) (string, bool)) (*Config, error) {
+	env := &environment{lookupEnv: lookupEnv}
+	cfg := &Config{
+		ListenAddr:              env.address("SWWAF_LISTEN_ADDR", ":8080"),
+		UpstreamURL:             env.appURL("SWWAF_UPSTREAM_URL", "http://127.0.0.1:8081"),
+		TrustedProxies:          env.netblocks("SWWAF_TRUSTED_PROXIES", privateRanges),
+		ClientRequestTimeout:    env.duration("SWWAF_CLIENT_REQUEST_TIMEOUT", "60s"),
+		ClientResponseTimeout:   env.duration("SWWAF_CLIENT_RESPONSE_TIMEOUT", "30m"),
+		UpstreamRequestTimeout:  env.duration("SWWAF_UPSTREAM_REQUEST_TIMEOUT", "60s"),
+		UpstreamResponseTimeout: env.duration("SWWAF_UPSTREAM_RESPONSE_TIMEOUT", "30m"),
+		RequestMaxBytes:         env.size("SWWAF_REQUEST_MAX_BYTES", "100M"),
+		ResponseMaxBytes:        env.size("SWWAF_RESPONSE_MAX_BYTES", "5G"),
+	}
+
+	if env.err != nil {
+		return nil, env.err
+	}
+
+	cfg.settings = env.settings
+
+	return cfg, nil
+}
+
+// privateRanges are the private address ranges, the default trusted
+// proxies.
+const privateRanges = "10.0.0.0/8,172.16.0.0/12,192.168.0.0/16"
+
+// LogValue makes a Config log as each setting's name with the value it was
+// given, or its default.
+func (c *Config) LogValue() slog.Value {
+	return slog.GroupValue(c.settings...)
+}
+
+// environment is where FromEnvironment reads the settings: it notes each
+// value for the log, and keeps the first error.
+type environment struct {
+	lookupEnv func(string) (string, bool)
+	settings  []slog.Attr
+	err       error
+}
+
+// value returns a setting's value, or its default when it is not set,
+// and notes it for the log.
+func (e *environment) value(name, defaultValue string) string {
+	value, ok := e.lookupEnv(name)
+	if !ok {
+		value = defaultValue
+	}
+
+	e.settings = append(e.settings, slog.String(name, value))
+
+	return value
+}
+
+// check keeps the first error, naming the setting it is about.
+func (e *environment) check(name string, err error) {
+	if err != nil && e.err == nil {
+		e.err = fmt.Errorf("%s: %w", name, err)
+	}
+}
+
+// address reads a setting that is an address to listen on.
+func (e *environment) address(name, defaultValue string) string {
+	address, err := parseListenAddr(e.value(name, defaultValue))
+	e.check(name, err)
+
+	return address
+}
+
+// appURL reads a setting that is the app's URL.
+func (e *environment) appURL(name, defaultValue string) *url.URL {
+	upstream, err := parseUpstreamURL(e.value(name, defaultValue))
+	e.check(name, err)
+
+	return upstream
+}
+
+// netblocks reads a setting that is a list of netblocks.
+func (e *environment) netblocks(name, defaultValue string) []netip.Prefix {
+	netblocks, err := parseNetblocks(e.value(name, defaultValue))
+	e.check(name, err)
+
+	return netblocks
+}
+
+// duration reads a setting that is a duration.
+func (e *environment) duration(name, defaultValue string) time.Duration {
+	duration, err := parseDuration(e.value(name, defaultValue))
+	e.check(name, err)
+
+	return duration
+}
+
+// size reads a setting that is a number of bytes.
+func (e *environment) size(name, defaultValue string) int64 {
+	size, err := parseSize(e.value(name, defaultValue))
+	e.check(name, err)
+
+	return size
+}
+
+// parseDuration reads a duration in Go's syntax, such as 90s or 15m, a
+// whole number of days such as 7d, or off.
+func parseDuration(value string) (time.Duration, error) {
+	if value == off {
+		return 0, nil
+	}
+
+	duration, err := durationOrDays(value)
+	if err != nil {
+		return 0, fmt.Errorf("%q %w", value, errNotDuration)
+	}
+
+	if duration <= 0 {
+		return 0, fmt.Errorf("%q %w", value, errNotPositive)
+	}
+
+	return duration, nil
+}
+
+// durationOrDays reads Go's duration syntax, or a whole number of days.
+func durationOrDays(value string) (time.Duration, error) {
+	days, isDays := strings.CutSuffix(value, "d")
+	if !isDays {
+		return time.ParseDuration(value)
+	}
+
+	n, err := strconv.ParseInt(days, 10, 64)
+	if err != nil || n < 0 || n > math.MaxInt64/int64(day) {
+		return 0, errNotDuration
+	}
+
+	return time.Duration(n) * day, nil
+}
+
+// parseSize reads a number of bytes with an optional K, M or G suffix, in
+// powers of 1024 (1K is 1024 bytes), or off.
+func parseSize(value string) (int64, error) {
+	if value == off {
+		return 0, nil
+	}
+
+	number, unit := splitUnit(value)
+
+	n, err := strconv.ParseInt(number, 10, 64)
+	if err != nil || n > math.MaxInt64/unit {
+		return 0, fmt.Errorf("%q %w", value, errNotSize)
+	}
+
+	if n <= 0 {
+		return 0, fmt.Errorf("%q %w", value, errNotPositive)
+	}
+
+	return n * unit, nil
+}
+
+// splitUnit splits a size into its number and the bytes its suffix
+// stands for.
+func splitUnit(value string) (string, int64) {
+	switch {
+	case strings.HasSuffix(value, "K"):
+		return strings.TrimSuffix(value, "K"), kibibyte
+	case strings.HasSuffix(value, "M"):
+		return strings.TrimSuffix(value, "M"), mebibyte
+	case strings.HasSuffix(value, "G"):
+		return strings.TrimSuffix(value, "G"), gibibyte
+	default:
+		return value, 1
+	}
+}
+
+// parseList splits a comma-separated list and trims the spaces around
+// each item. An empty value is an empty list.
+func parseList(value string) ([]string, error) {
+	if strings.TrimSpace(value) == "" {
+		return []string{}, nil
+	}
+
+	items := strings.Split(value, ",")
+	for i, item := range items {
+		items[i] = strings.TrimSpace(item)
+		if items[i] == "" {
+			return nil, fmt.Errorf("%q %w", value, errEmptyItem)
+		}
+	}
+
+	return items, nil
+}
+
+// parseNetblocks reads a comma-separated list of netblocks.
+func parseNetblocks(value string) ([]netip.Prefix, error) {
+	items, err := parseList(value)
+	if err != nil {
+		return nil, err
+	}
+
+	netblocks := make([]netip.Prefix, 0, len(items))
+
+	for _, item := range items {
+		netblock, err := parseNetblock(item)
+		if err != nil {
+			return nil, err
+		}
+
+		netblocks = append(netblocks, netblock)
+	}
+
+	return netblocks, nil
+}
+
+// parseNetblock reads a netblock in CIDR form, such as 10.0.0.0/8. A bare
+// address is a netblock of that address alone, a /32 or a /128.
+func parseNetblock(value string) (netip.Prefix, error) {
+	if strings.Contains(value, "/") {
+		netblock, err := netip.ParsePrefix(value)
+		if err != nil {
+			return netip.Prefix{}, fmt.Errorf("%q %w", value, errNotNetblock)
+		}
+
+		return netblock.Masked(), nil
+	}
+
+	addr, err := netip.ParseAddr(value)
+	if err != nil || addr.Zone() != "" {
+		return netip.Prefix{}, fmt.Errorf("%q %w", value, errNotNetblock)
+	}
+
+	return netip.PrefixFrom(addr, addr.BitLen()), nil
+}
+
+// parseListenAddr checks an address to listen on: an optional host and a
+// port number.
+func parseListenAddr(value string) (string, error) {
+	_, port, err := net.SplitHostPort(value)
+	if err != nil {
+		return "", fmt.Errorf("%q %w", value, errNotListenAddr)
+	}
+
+	_, err = strconv.ParseUint(port, 10, 16)
+	if err != nil {
+		return "", fmt.Errorf("%q %w", value, errNotListenAddr)
+	}
+
+	return value, nil
+}
+
+// parseUpstreamURL reads the app's URL: http or https, a host and an
+// optional port from 1 to 65535, and nothing else, since the request's
+// own path and query go to the app unchanged.
+func parseUpstreamURL(value string) (*url.URL, error) {
+	upstream, err := url.Parse(value)
+	if err != nil {
+		return nil, fmt.Errorf("%q %w", value, errNotUpstreamURL)
+	}
+
+	onlySchemeAndHost := (upstream.Scheme == "http" || upstream.Scheme == "https") &&
+		upstream.Hostname() != "" && upstream.User == nil && upstream.Opaque == "" &&
+		(upstream.Path == "" || upstream.Path == "/") &&
+		upstream.RawQuery == "" && upstream.Fragment == ""
+	if !onlySchemeAndHost {
+		return nil, fmt.Errorf("%q %w", value, errNotUpstreamURL)
+	}
+
+	if upstream.Port() != "" {
+		port, err := strconv.ParseUint(upstream.Port(), 10, 16)
+		if err != nil || port == 0 {
+			return nil, fmt.Errorf("%q %w", value, errNotUpstreamURL)
+		}
+	}
+
+	return upstream, nil
+}

internal/config/config_test.go

@@ -0,0 +1,244 @@
+package config_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"log/slog"
+	"maps"
+	"net/netip"
+	"slices"
+	"strings"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/config"
+)
+
+// The settings, by name.
+const (
+	listenAddr              = "SWWAF_LISTEN_ADDR"
+	upstreamURL             = "SWWAF_UPSTREAM_URL"
+	trustedProxies          = "SWWAF_TRUSTED_PROXIES"
+	clientRequestTimeout    = "SWWAF_CLIENT_REQUEST_TIMEOUT"
+	clientResponseTimeout   = "SWWAF_CLIENT_RESPONSE_TIMEOUT"
+	upstreamRequestTimeout  = "SWWAF_UPSTREAM_REQUEST_TIMEOUT"
+	upstreamResponseTimeout = "SWWAF_UPSTREAM_RESPONSE_TIMEOUT"
+	requestMaxBytes         = "SWWAF_REQUEST_MAX_BYTES"
+	responseMaxBytes        = "SWWAF_RESPONSE_MAX_BYTES"
+)
+
+// off switches a timeout or a size limit off.
+const off = "off"
+
+// environment is a set of environment variables, for FromEnvironment.
+type environment map[string]string
+
+// lookupEnv reads one of the variables, as os.LookupEnv does.
+func (e environment) lookupEnv(name string) (string, bool) {
+	value, ok := e[name]
+
+	return value, ok
+}
+
+// fromEnvironment reads the settings from env, which must be valid.
+func fromEnvironment(t *testing.T, env environment) *config.Config {
+	t.Helper()
+
+	cfg, err := config.FromEnvironment(env.lookupEnv)
+	if err != nil {
+		t.Fatalf("settings %v: %v", env, err)
+	}
+
+	return cfg
+}
+
+func TestDefaults(t *testing.T) {
+	t.Parallel()
+
+	cfg := fromEnvironment(t, environment{})
+
+	wantSettings(t, cfg, config.Config{
+		ListenAddr:              ":8080",
+		ClientRequestTimeout:    time.Minute,
+		ClientResponseTimeout:   30 * time.Minute,
+		UpstreamRequestTimeout:  time.Minute,
+		UpstreamResponseTimeout: 30 * time.Minute,
+		RequestMaxBytes:         100 << 20,
+		ResponseMaxBytes:        5 << 30,
+	})
+
+	if cfg.UpstreamURL.String() != "http://127.0.0.1:8081" {
+		t.Errorf("%s is %s", upstreamURL, cfg.UpstreamURL)
+	}
+
+	wantNetblocks(t, cfg.TrustedProxies,
+		"10.0.0.0/8", "172.16.0.0/12", "192.168.0.0/16")
+}
+
+func TestValuesAsSet(t *testing.T) {
+	t.Parallel()
+
+	cfg := fromEnvironment(t, environment{
+		listenAddr:              "127.0.0.1:9000",
+		upstreamURL:             "https://app.internal:8443/",
+		trustedProxies:          " 192.0.2.1, 10.1.2.3/8 ,2001:db8::/32",
+		clientRequestTimeout:    "90s",
+		clientResponseTimeout:   "7d",
+		upstreamRequestTimeout:  "1h30m",
+		upstreamResponseTimeout: off,
+		requestMaxBytes:         "512K",
+		responseMaxBytes:        "1234",
+	})
+
+	wantSettings(t, cfg, config.Config{
+		ListenAddr:              "127.0.0.1:9000",
+		ClientRequestTimeout:    90 * time.Second,
+		ClientResponseTimeout:   7 * 24 * time.Hour,
+		UpstreamRequestTimeout:  90 * time.Minute,
+		UpstreamResponseTimeout: 0,
+		RequestMaxBytes:         512 << 10,
+		ResponseMaxBytes:        1234,
+	})
+
+	if cfg.UpstreamURL.String() != "https://app.internal:8443/" {
+		t.Errorf("%s is %s", upstreamURL, cfg.UpstreamURL)
+	}
+
+	wantNetblocks(t, cfg.TrustedProxies, "192.0.2.1/32", "10.0.0.0/8", "2001:db8::/32")
+}
+
+func TestSizesAndOff(t *testing.T) {
+	t.Parallel()
+
+	cfg := fromEnvironment(t, environment{
+		requestMaxBytes:      "3G",
+		responseMaxBytes:     off,
+		clientRequestTimeout: off,
+	})
+
+	if cfg.RequestMaxBytes != 3<<30 || cfg.ResponseMaxBytes != 0 ||
+		cfg.ClientRequestTimeout != 0 {
+		t.Errorf("3G, off and off read as %d, %d and %s",
+			cfg.RequestMaxBytes, cfg.ResponseMaxBytes, cfg.ClientRequestTimeout)
+	}
+}
+
+func TestTrustedProxiesSetButEmptyTrustNothing(t *testing.T) {
+	t.Parallel()
+
+	cfg := fromEnvironment(t, environment{trustedProxies: ""})
+	if len(cfg.TrustedProxies) != 0 {
+		t.Errorf("trusted proxies %v, want none", cfg.TrustedProxies)
+	}
+}
+
+func TestInvalidValueStopsTheStart(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range []struct{ name, value string }{
+		{listenAddr, "8080"},
+		{listenAddr, ":http"},
+		{listenAddr, ":65536"},
+		{upstreamURL, "127.0.0.1:8081"},
+		{upstreamURL, "ftp://127.0.0.1:8081"},
+		{upstreamURL, "http://"},
+		{upstreamURL, "http://:8081"},
+		{upstreamURL, "http://127.0.0.1:0"},
+		{upstreamURL, "http://127.0.0.1:99999"},
+		{upstreamURL, "http://127.0.0.1:8081/app"},
+		{upstreamURL, "http://127.0.0.1:8081/?a=1"},
+		{upstreamURL, "http://user:secret@127.0.0.1:8081"},
+		{trustedProxies, "10.0.0.0/33"},
+		{trustedProxies, "traefik"},
+		{trustedProxies, "10.0.0.0/8,,192.168.0.0/16"},
+		{trustedProxies, "fe80::1%eth0"},
+		{clientRequestTimeout, "60"},
+		{clientRequestTimeout, ""},
+		{clientResponseTimeout, "1y"},
+		{upstreamRequestTimeout, "-1s"},
+		{upstreamResponseTimeout, "0s"},
+		{upstreamResponseTimeout, "1.5d"},
+		{requestMaxBytes, "100MB"},
+		{requestMaxBytes, "100m"},
+		{requestMaxBytes, "1.5M"},
+		{responseMaxBytes, "0"},
+		{responseMaxBytes, "-5"},
+		{responseMaxBytes, "99999999999G"},
+	} {
+		t.Run(tc.name+"="+tc.value, func(t *testing.T) {
+			t.Parallel()
+
+			_, err := config.FromEnvironment(environment{tc.name: tc.value}.lookupEnv)
+			if err == nil {
+				t.Fatalf("%s=%q was accepted", tc.name, tc.value)
+			}
+
+			if !strings.HasPrefix(err.Error(), tc.name+": ") {
+				t.Errorf("error %q does not name %s", err, tc.name)
+			}
+		})
+	}
+}
+
+func TestLogsEachSettingWithItsValue(t *testing.T) {
+	t.Parallel()
+
+	cfg := fromEnvironment(t, environment{clientRequestTimeout: "45s"})
+
+	var out bytes.Buffer
+
+	slog.New(slog.NewJSONHandler(&out, nil)).Info("starting", "settings", cfg)
+
+	var line struct {
+		Settings map[string]string `json:"settings"`
+	}
+
+	err := json.Unmarshal(out.Bytes(), &line)
+	if err != nil {
+		t.Fatalf("decode %s: %v", out.Bytes(), err)
+	}
+
+	want := map[string]string{
+		listenAddr:              ":8080",
+		upstreamURL:             "http://127.0.0.1:8081",
+		trustedProxies:          "10.0.0.0/8,172.16.0.0/12,192.168.0.0/16",
+		clientRequestTimeout:    "45s",
+		clientResponseTimeout:   "30m",
+		upstreamRequestTimeout:  "60s",
+		upstreamResponseTimeout: "30m",
+		requestMaxBytes:         "100M",
+		responseMaxBytes:        "5G",
+	}
+	if !maps.Equal(line.Settings, want) {
+		t.Errorf("logged settings\n%v\nwant\n%v", line.Settings, want)
+	}
+}
+
+// wantSettings checks the settings that are plain values.
+func wantSettings(t *testing.T, got *config.Config, want config.Config) {
+	t.Helper()
+
+	if got.ListenAddr != want.ListenAddr ||
+		got.ClientRequestTimeout != want.ClientRequestTimeout ||
+		got.ClientResponseTimeout != want.ClientResponseTimeout ||
+		got.UpstreamRequestTimeout != want.UpstreamRequestTimeout ||
+		got.UpstreamResponseTimeout != want.UpstreamResponseTimeout ||
+		got.RequestMaxBytes != want.RequestMaxBytes ||
+		got.ResponseMaxBytes != want.ResponseMaxBytes {
+		t.Errorf("settings\n%+v\nwant\n%+v", got, want)
+	}
+}
+
+// wantNetblocks checks a list of netblocks.
+func wantNetblocks(t *testing.T, got []netip.Prefix, want ...string) {
+	t.Helper()
+
+	gotText := make([]string, 0, len(got))
+	for _, netblock := range got {
+		gotText = append(gotText, netblock.String())
+	}
+
+	if !slices.Equal(gotText, want) {
+		t.Errorf("netblocks %v, want %v", gotText, want)
+	}
+}

internal/proxy/bodies.go

@@ -0,0 +1,171 @@
+package proxy
+
+import (
+	"errors"
+	"io"
+	"net/http"
+	"sync/atomic"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// errResponseTooLarge ends an app's response body that is longer than
+// SWWAF_RESPONSE_MAX_BYTES.
+var errResponseTooLarge = errors.New(
+	"the response body is over SWWAF_RESPONSE_MAX_BYTES")
+
+// requestBody is the client's request body on its way to the app. The
+// transport reads it on a goroutine of its own.
+type requestBody struct {
+	// body is the client's body, ending in an *http.MaxBytesError past
+	// SWWAF_REQUEST_MAX_BYTES.
+	body io.ReadCloser
+	rq   *request
+	// waiting is true while a Read waits for the client to send more.
+	waiting atomic.Bool
+	// received is true once the client has sent the whole body.
+	received atomic.Bool
+	// bytes is how much of the body has been read.
+	bytes atomic.Int64
+}
+
+// Read reads from the client's body.
+func (b *requestBody) Read(p []byte) (int, error) {
+	b.waiting.Store(true)
+	n, err := b.body.Read(p)
+	b.waiting.Store(false)
+	b.bytes.Add(int64(n))
+
+	var tooLarge *http.MaxBytesError
+
+	switch {
+	case errors.Is(err, io.EOF):
+		b.received.Store(true)
+		b.rq.bodyReceived()
+	case errors.As(err, &tooLarge):
+		b.rq.refuse(refusal{
+			status: http.StatusRequestEntityTooLarge,
+			action: requestlog.ActionTooLarge,
+		})
+	}
+
+	return n, err
+}
+
+// Close closes the client's body.
+func (b *requestBody) Close() error {
+	return b.body.Close()
+}
+
+// responseBody is the app's response body on its way to the client.
+type responseBody struct {
+	// body is the app's body, ending in an *http.MaxBytesError past
+	// SWWAF_RESPONSE_MAX_BYTES.
+	body io.ReadCloser
+	rq   *request
+}
+
+// Read reads from the app's body.
+func (b *responseBody) Read(p []byte) (int, error) {
+	n, err := b.body.Read(p)
+	if err == nil {
+		return n, nil
+	}
+
+	var tooLarge *http.MaxBytesError
+
+	switch {
+	case errors.Is(err, io.EOF):
+		b.rq.responseReceived()
+	case errors.As(err, &tooLarge):
+		b.rq.refuse(refusal{
+			status: http.StatusBadGateway,
+			action: requestlog.ActionTooLarge,
+		})
+
+		return n, errResponseTooLarge
+	case b.rq.in.Context().Err() == nil:
+		// The answer broke off, not because the client went away. If a
+		// timeout cut it, that refusal came first and is the one kept.
+		b.rq.refuse(refusal{
+			status: http.StatusBadGateway,
+			action: requestlog.ActionUpstreamError,
+		})
+	}
+
+	return n, err
+}
+
+// Close closes the app's body.
+func (b *responseBody) Close() error {
+	return b.body.Close()
+}
+
+// limitBody returns body, cut off with an *http.MaxBytesError after
+// maxBytes, or unchanged if maxBytes is zero, which is off.
+func limitBody(body io.ReadCloser, maxBytes int64) io.ReadCloser {
+	if maxBytes == 0 {
+		return body
+	}
+
+	// Without a ResponseWriter, MaxBytesReader only counts and cuts off.
+	return http.MaxBytesReader(nil, body, maxBytes)
+}
+
+// responseWriter is the response to the client. It notes the status and
+// size for the log line, and the first error writing to the client.
+type responseWriter struct {
+	http.ResponseWriter
+
+	// status is the final status sent, or zero before one is.
+	status int
+	bytes  int64
+	err    error
+}
+
+// WriteHeader sends the status and headers. An informational 1xx status
+// is passed on and the final status still comes later.
+func (w *responseWriter) WriteHeader(status int) {
+	if status >= http.StatusOK && w.status == 0 {
+		w.status = status
+	}
+
+	w.ResponseWriter.WriteHeader(status)
+}
+
+// Write sends part of the body.
+func (w *responseWriter) Write(p []byte) (int, error) {
+	if w.status == 0 {
+		w.status = http.StatusOK
+	}
+
+	n, err := w.ResponseWriter.Write(p)
+	w.bytes += int64(n)
+	w.noteError(err)
+
+	return n, err
+}
+
+// FlushError sends what has been written so far.
+// http.ResponseController calls it, as ReverseProxy does after each
+// write.
+func (w *responseWriter) FlushError() error {
+	err := http.NewResponseController(w.ResponseWriter).Flush()
+	w.noteError(err)
+
+	return err
+}
+
+// Unwrap lets http.ResponseController reach net/http's own
+// ResponseWriter, which is how ReverseProxy takes over the connection of
+// an upgraded request.
+func (w *responseWriter) Unwrap() http.ResponseWriter {
+	return w.ResponseWriter
+}
+
+// noteError keeps the first error writing to the client.
+func (w *responseWriter) noteError(err error) {
+	if w.err == nil {
+		w.err = err
+	}
+}

internal/proxy/client.go

@@ -0,0 +1,101 @@
+package proxy
+
+import (
+	"net/http"
+	"net/netip"
+	"slices"
+	"strings"
+)
+
+// peerAddress is the address of the request's TCP peer, normally traefik.
+func peerAddress(r *http.Request) netip.Addr {
+	addrPort, err := netip.ParseAddrPort(r.RemoteAddr)
+	if err != nil {
+		return netip.Addr{}
+	}
+
+	return addrPort.Addr().Unmap()
+}
+
+// clientAddress works out who the client is. A peer outside the trusted
+// proxies is the client, and what it says in X-Forwarded-For is ignored.
+// For a peer inside them, X-Forwarded-For is read from the right, and the
+// first address outside them is the client; if every address in it is
+// inside, the leftmost is, and with no header, the peer. An entry that is
+// not an address ends the reading, since nothing to its left can be
+// believed.
+func clientAddress(
+	peer netip.Addr, forwardedFor []string, trusted []netip.Prefix,
+) netip.Addr {
+	client := peer
+	if !isInside(peer, trusted) {
+		return client
+	}
+
+	entries := strings.Split(strings.Join(forwardedFor, ","), ",")
+	for _, entry := range slices.Backward(entries) {
+		addr, err := netip.ParseAddr(strings.TrimSpace(entry))
+		if err != nil {
+			break
+		}
+
+		client = addr.Unmap()
+		if !isInside(client, trusted) {
+			break
+		}
+	}
+
+	return client
+}
+
+// isInside reports whether addr is in one of the netblocks.
+func isInside(addr netip.Addr, netblocks []netip.Prefix) bool {
+	return slices.ContainsFunc(netblocks, func(netblock netip.Prefix) bool {
+		return netblock.Contains(addr)
+	})
+}
+
+// setForwardedHeaders sets the headers in which the app learns about the
+// client, so that it sees what it would see from traefik directly. A
+// trusted proxy's forwarded headers pass on, with the proxy's own address
+// added to X-Forwarded-For. Those of any other peer are its own claims and
+// are replaced: X-Forwarded-For names the peer, X-Forwarded-Host the host
+// it asked for, and X-Forwarded-Proto plain http, which is how it reached
+// smallwebwaf.
+func setForwardedHeaders(in, out *http.Request, peer netip.Addr, trusted bool) {
+	forwardedFor := peer.String()
+
+	if trusted {
+		// ReverseProxy removes these from out before Rewrite.
+		for _, name := range []string{"Forwarded", "X-Forwarded-Host", "X-Forwarded-Proto"} {
+			values, ok := in.Header[name]
+			if ok {
+				out.Header[name] = values
+			}
+		}
+
+		prior := in.Header.Values("X-Forwarded-For")
+		if len(prior) > 0 {
+			forwardedFor = strings.Join(prior, ", ") + ", " + forwardedFor
+		}
+
+		out.Header.Set("X-Forwarded-For", forwardedFor)
+
+		return
+	}
+
+	// ReverseProxy has removed Forwarded and the three set below; these
+	// are the other headers in which traefik tells the app about the
+	// client and its request.
+	for _, name := range []string{
+		"X-Forwarded-Port", "X-Forwarded-Server", "X-Forwarded-Uri",
+		"X-Forwarded-Method", "X-Forwarded-Prefix", "X-Forwarded-Tls-Client-Cert",
+		"X-Forwarded-Tls-Client-Cert-Info", "X-Real-Ip",
+	} {
+		out.Header.Del(name)
+	}
+
+	out.Header.Set("X-Forwarded-For", forwardedFor)
+	out.Header.Set("X-Forwarded-Host", in.Host)
+	out.Header.Set("X-Forwarded-Proto", "http")
+}

internal/proxy/client_test.go

@@ -0,0 +1,161 @@
+package proxy_test
+
+import (
+	"encoding/json"
+	"net/http"
+	"testing"
+)
+
+const (
+	// trustLocalhost trusts the address every test connects from, and a
+	// network for proxies in front of it.
+	trustLocalhost = localhost + "/32,10.0.0.0/8"
+	// appHost is the host every test asks for.
+	appHost = "app.example"
+	// client is the client's address, as a proxy names it.
+	client = "203.0.113.9"
+	// forwardedFor is the header that lists the client and its proxies.
+	forwardedFor = "X-Forwarded-For"
+	// secure is the scheme a client reached traefik with.
+	secure = "https"
+)
+
+// appHeaders is what the app tells about the headers it received.
+type appHeaders struct {
+	Host           string `json:"host"`
+	ForwardedFor   string `json:"forwardedFor"`
+	ForwardedHost  string `json:"forwardedHost"`
+	ForwardedProto string `json:"forwardedProto"`
+	RealIP         string `json:"realIp"`
+}
+
+// clientAddressCase is a request and what smallwebwaf makes of it.
+type clientAddressCase struct {
+	name       string
+	env        map[string]string
+	header     http.Header
+	wantClient string
+	wantApp    appHeaders
+}
+
+func TestClientAddressAndForwardedHeaders(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range clientAddressCases() {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			got, line := requestWithHeaders(t, tc.env, tc.header)
+
+			tc.wantApp.Host = appHost
+			if got != tc.wantApp {
+				t.Errorf("app received %+v, want %+v", got, tc.wantApp)
+			}
+
+			if line.ClientIP != tc.wantClient || line.PeerIP != localhost {
+				t.Errorf("log line has client_ip %q and peer_ip %q, want %q and %q",
+					line.ClientIP, line.PeerIP, tc.wantClient, localhost)
+			}
+		})
+	}
+}
+
+// clientAddressCases are the requests TestClientAddressAndForwardedHeaders
+// sends, from 127.0.0.1, which the default trusted proxies leave out.
+func clientAddressCases() []clientAddressCase {
+	trusted := map[string]string{trustedProxies: trustLocalhost}
+	forged := http.Header{
+		forwardedFor:        {client},
+		"X-Forwarded-Host":  {"forged.example"},
+		"X-Forwarded-Proto": {secure},
+		"X-Real-Ip":         {client},
+	}
+	replaced := appHeaders{
+		ForwardedFor: localhost, ForwardedHost: appHost, ForwardedProto: "http",
+	}
+
+	return []clientAddressCase{{
+		name: "a peer outside the trusted proxies is the client, " +
+			"and its forwarded headers are replaced",
+		header: forged, wantClient: localhost, wantApp: replaced,
+	}, {
+		name:   "set but empty, the trusted proxies trust nothing",
+		env:    map[string]string{trustedProxies: ""},
+		header: forged, wantClient: localhost, wantApp: replaced,
+	}, {
+		name: "behind a trusted peer, the client is the first address " +
+			"outside the trusted proxies from the right",
+		env: trusted,
+		header: http.Header{
+			forwardedFor:        {"198.51.100.7, " + client + ", 10.0.0.2"},
+			"X-Forwarded-Host":  {appHost},
+			"X-Forwarded-Proto": {secure},
+			"X-Real-Ip":         {client},
+		},
+		wantClient: client,
+		wantApp: appHeaders{
+			ForwardedFor:  "198.51.100.7, " + client + ", 10.0.0.2, " + localhost,
+			ForwardedHost: appHost, ForwardedProto: secure, RealIP: client,
+		},
+	}, {
+		name:       "when every address is a trusted proxy, the leftmost is the client",
+		env:        trusted,
+		header:     http.Header{forwardedFor: {"10.0.0.5, 10.0.0.2"}},
+		wantClient: "10.0.0.5",
+		wantApp:    appHeaders{ForwardedFor: "10.0.0.5, 10.0.0.2, " + localhost},
+	}, {
+		name:       "with no header, a trusted peer is the client",
+		env:        trusted,
+		wantClient: localhost,
+		wantApp:    appHeaders{ForwardedFor: localhost},
+	}, {
+		name:       "an entry that is not an address ends the reading",
+		env:        trusted,
+		header:     http.Header{forwardedFor: {client + ", unknown, 10.0.0.2"}},
+		wantClient: "10.0.0.2",
+		wantApp: appHeaders{
+			ForwardedFor: client + ", unknown, 10.0.0.2, " + localhost,
+		},
+	}, {
+		name:       "several header lines are read as one list",
+		env:        trusted,
+		header:     http.Header{forwardedFor: {"2001:db8::7", "10.0.0.2"}},
+		wantClient: "2001:db8::7",
+		wantApp:    appHeaders{ForwardedFor: "2001:db8::7, 10.0.0.2, " + localhost},
+	}}
+}
+
+// requestWithHeaders sends a request for appHost with header through
+// smallwebwaf, with the settings in env, and returns the headers the app
+// received and the request's log line.
+func requestWithHeaders(
+	t *testing.T, env map[string]string, header http.Header,
+) (appHeaders, logLine) {
+	t.Helper()
+
+	app := startApp(t, func(w http.ResponseWriter, r *http.Request) {
+		_ = json.NewEncoder(w).Encode(appHeaders{
+			Host:           r.Host,
+			ForwardedFor:   r.Header.Get(forwardedFor),
+			ForwardedHost:  r.Header.Get("X-Forwarded-Host"),
+			ForwardedProto: r.Header.Get("X-Forwarded-Proto"),
+			RealIP:         r.Header.Get("X-Real-Ip"),
+		})
+	})
+	addr, out := startProxy(t, app.URL, env)
+
+	req := newRequest(t, http.MethodGet, addr, "/", http.NoBody)
+	req.Host = appHost
+	req.Header = header.Clone()
+
+	answered := do(t, req)
+
+	var got appHeaders
+
+	err := json.Unmarshal(answered.body, &got)
+	if err != nil {
+		t.Fatalf("decode the app's answer %q: %v", answered.body, err)
+	}
+
+	return got, out.requestLine(t)
+}

internal/proxy/limits_test.go

@@ -0,0 +1,146 @@
+package proxy_test
+
+import (
+	"bytes"
+	"errors"
+	"io"
+	"net/http"
+	"strconv"
+	"sync/atomic"
+	"testing"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// sizeLimit is the size limit the tests set, 1K as a setting.
+const (
+	sizeLimit        = 1 << 10
+	sizeLimitSetting = "1K"
+)
+
+func TestRequestBodyLimit(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range []struct {
+		name string
+		size int
+		// announced sends the size in Content-Length; otherwise the body
+		// is sent in chunks with no length given.
+		announced bool
+		want      int
+		action    string
+		// refusedBeforeApp is a refusal before anything reaches the app.
+		// A body over the limit with no length given has already partly
+		// reached the app when it is refused.
+		refusedBeforeApp bool
+	}{
+		{"announced, over the limit", 2 * sizeLimit, true,
+			http.StatusRequestEntityTooLarge, requestlog.ActionTooLarge, true},
+		{"announced, at the limit", sizeLimit, true,
+			http.StatusOK, requestlog.ActionForward, false},
+		{"not announced, over the limit", 4 * sizeLimit, false,
+			http.StatusRequestEntityTooLarge, requestlog.ActionTooLarge, false},
+		{"not announced, at the limit", sizeLimit, false,
+			http.StatusOK, requestlog.ActionForward, false},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			var calls atomic.Int32
+
+			app := startApp(t, func(_ http.ResponseWriter, r *http.Request) {
+				calls.Add(1)
+
+				_, _ = io.Copy(io.Discard, r.Body)
+			})
+			addr, out := startProxy(t, app.URL, map[string]string{
+				requestMaxBytes: sizeLimitSetting,
+			})
+
+			var body io.Reader = bytes.NewReader(make([]byte, tc.size))
+			if !tc.announced {
+				body = io.MultiReader(body) // hides the length
+			}
+
+			wantStatus(t, do(t, newRequest(t, http.MethodPost, addr, "/upload", body)),
+				tc.want)
+			wantLine(t, out.requestLine(t), tc.want, tc.action)
+
+			if tc.refusedBeforeApp && calls.Load() != 0 {
+				t.Errorf("the app was called %d times, want never", calls.Load())
+			}
+		})
+	}
+}
+
+func TestResponseBodyLimit(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range []struct {
+		name string
+		size int
+		// announced sends the size in Content-Length; otherwise the body
+		// is sent in chunks with no length given.
+		announced bool
+		want      int
+		action    string
+		// received is how much of a body the client gets, and cutOff
+		// whether the connection is then cut.
+		received int
+		cutOff   bool
+	}{
+		{"announced, over the limit", 2 * sizeLimit, true, http.StatusBadGateway,
+			requestlog.ActionTooLarge, len("Bad Gateway\n"), false},
+		{"announced, at the limit", sizeLimit, true, http.StatusOK,
+			requestlog.ActionForward, sizeLimit, false},
+		{"not announced, over the limit", 4 * sizeLimit, false, http.StatusOK,
+			requestlog.ActionTooLarge, sizeLimit, true},
+		{"not announced, at the limit", sizeLimit, false, http.StatusOK,
+			requestlog.ActionForward, sizeLimit, false},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			app := startApp(t, func(w http.ResponseWriter, _ *http.Request) {
+				answerWithSize(w, tc.size, tc.announced)
+			})
+			addr, out := startProxy(t, app.URL, map[string]string{
+				responseMaxBytes: sizeLimitSetting,
+			})
+
+			got := get(t, addr, "/download")
+			wantStatus(t, got, tc.want)
+
+			if len(got.body) != tc.received ||
+				errors.Is(got.err, io.ErrUnexpectedEOF) != tc.cutOff {
+				t.Errorf("client got %d bytes (%v), want %d",
+					len(got.body), got.err, tc.received)
+			}
+
+			line := out.requestLine(t)
+			wantLine(t, line, tc.want, tc.action)
+
+			if line.UpstreamStatus != http.StatusOK {
+				t.Errorf("log line has upstream_status %d", line.UpstreamStatus)
+			}
+		})
+	}
+}
+
+// answerWithSize answers with a body of size bytes, announced in
+// Content-Length or sent in chunks with no length given.
+func answerWithSize(w http.ResponseWriter, size int, announced bool) {
+	body := make([]byte, size)
+	if announced {
+		w.Header().Set("Content-Length", strconv.Itoa(size))
+		_, _ = w.Write(body)
+
+		return
+	}
+
+	// Sending part of it before the end keeps Go's server from working
+	// out the length.
+	_, _ = w.Write(body[:size/2])
+	_ = http.NewResponseController(w).Flush()
+	_, _ = w.Write(body[size/2:])
+}

internal/proxy/passthrough_test.go

@@ -0,0 +1,426 @@
+package proxy_test
+
+import (
+	"bufio"
+	"bytes"
+	"errors"
+	"io"
+	"net"
+	"net/http"
+	"slices"
+	"strings"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/config"
+	"sneak.berlin/go/smallwebwaf/internal/proxy"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// A request target with an escaped slash and space in its path, and a
+// query with a parameter ReverseProxy cannot parse.
+const (
+	rawPath  = "/some%2Fpath/with%20space"
+	rawQuery = "b=2&a=1&bad=%zz;x"
+)
+
+// chunkSize is the size of each part of a body a test sends in parts.
+const chunkSize = 1 << 10
+
+var errNotStreamed = errors.New("the first part never reached the app")
+
+// appSaw is what the app received.
+type appSaw struct {
+	method string
+	target string
+	header http.Header
+	body   []byte
+}
+
+func TestPassesRequestAndAnswerUnchanged(t *testing.T) {
+	t.Parallel()
+
+	requestBody := bytes.Repeat([]byte("request body "), 8000)
+	answerBody := bytes.Repeat([]byte("answer body "), 8000)
+	saw := make(chan appSaw, 1)
+
+	app := startApp(t, func(w http.ResponseWriter, r *http.Request) {
+		body, _ := io.ReadAll(r.Body)
+		saw <- appSaw{r.Method, r.RequestURI, r.Header.Clone(), body}
+
+		w.Header().Set("X-App", "yes")
+		w.Header().Add("Set-Cookie", "a=1")
+		w.Header().Add("Set-Cookie", "b=2")
+		w.WriteHeader(http.StatusTeapot)
+		_, _ = w.Write(answerBody)
+	})
+	addr, out := startProxy(t, app.URL, nil)
+
+	req := newRequest(t, http.MethodPatch, addr, rawPath+"?"+rawQuery,
+		bytes.NewReader(requestBody))
+	req.Header.Add("X-Test", "one")
+	req.Header.Add("X-Test", "two")
+	req.Header.Set("User-Agent", "test-agent")
+
+	got := do(t, req)
+
+	wantAppSaw(t, <-saw, requestBody)
+	wantAnswer(t, got, answerBody)
+
+	line := out.requestLine(t)
+	wantLine(t, line, http.StatusTeapot, requestlog.ActionForward)
+	wantRequestFields(t, line, addr, len(requestBody), len(answerBody))
+}
+
+// wantAppSaw checks that the app received the test's request unchanged.
+func wantAppSaw(t *testing.T, saw appSaw, body []byte) {
+	t.Helper()
+
+	if saw.method != http.MethodPatch || saw.target != rawPath+"?"+rawQuery {
+		t.Errorf("app saw %s %s, want %s %s", saw.method, saw.target,
+			http.MethodPatch, rawPath+"?"+rawQuery)
+	}
+
+	if !slices.Equal(saw.header.Values("X-Test"), []string{"one", "two"}) {
+		t.Errorf("app saw X-Test %q", saw.header.Values("X-Test"))
+	}
+
+	if saw.header.Get("User-Agent") != "test-agent" {
+		t.Errorf("app saw User-Agent %q", saw.header.Get("User-Agent"))
+	}
+
+	if !bytes.Equal(saw.body, body) {
+		t.Errorf("app saw a body of %d bytes, want the %d sent",
+			len(saw.body), len(body))
+	}
+}
+
+// wantAnswer checks that the client received the app's answer unchanged.
+func wantAnswer(t *testing.T, got answer, body []byte) {
+	t.Helper()
+
+	wantStatus(t, got, http.StatusTeapot)
+
+	if got.header.Get("X-App") != "yes" {
+		t.Errorf("client got X-App %q", got.header.Get("X-App"))
+	}
+
+	if !slices.Equal(got.header.Values("Set-Cookie"), []string{"a=1", "b=2"}) {
+		t.Errorf("client got Set-Cookie %q", got.header.Values("Set-Cookie"))
+	}
+
+	if got.err != nil || !bytes.Equal(got.body, body) {
+		t.Errorf("client got %d bytes (%v), want the %d the app sent",
+			len(got.body), got.err, len(body))
+	}
+}
+
+// wantRequestFields checks the log line's fields about the request.
+func wantRequestFields(t *testing.T, line logLine, host string, sent, received int) {
+	t.Helper()
+
+	want := requestlog.Line{
+		Type: "request", Time: line.Time, ClientIP: localhost, PeerIP: localhost,
+		Method: http.MethodPatch, Host: host, Path: rawPath, Query: rawQuery,
+		Protocol: "HTTP/1.1", Status: http.StatusTeapot,
+		UpstreamStatus: http.StatusTeapot, RequestBytes: int64(sent),
+		ResponseBytes: int64(received), UserAgent: "test-agent",
+		Action: requestlog.ActionForward, DurationTotal: line.DurationTotal,
+		DurationUpstreamTotal: line.DurationUpstreamTotal,
+	}
+	if line.Line != want {
+		t.Errorf("log line\n%+v\nwant\n%+v", line.Line, want)
+	}
+
+	_, err := time.Parse(time.RFC3339, line.Time)
+	if err != nil || line.DurationTotal <= 0 || line.DurationUpstreamTotal <= 0 {
+		t.Errorf("log line has time %q and durations %v and %v",
+			line.Time, line.DurationTotal, line.DurationUpstreamTotal)
+	}
+}
+
+func TestStreamsTheRequestBody(t *testing.T) {
+	t.Parallel()
+
+	chunk := bytes.Repeat([]byte("x"), chunkSize)
+	firstArrived := make(chan struct{})
+
+	app := startApp(t, func(w http.ResponseWriter, r *http.Request) {
+		first := make([]byte, len(chunk))
+
+		_, err := io.ReadFull(r.Body, first)
+		if err != nil {
+			return
+		}
+
+		close(firstArrived)
+
+		rest, _ := io.ReadAll(r.Body)
+		_, _ = w.Write(rest)
+	})
+	addr, _ := startProxy(t, app.URL, nil)
+
+	body, writer := io.Pipe()
+
+	go func() {
+		_, _ = writer.Write(chunk)
+
+		select {
+		case <-firstArrived:
+			_, _ = writer.Write(chunk)
+			_ = writer.Close()
+		case <-time.After(waitLimit):
+			_ = writer.CloseWithError(errNotStreamed)
+		}
+	}()
+
+	got := do(t, newRequest(t, http.MethodPost, addr, "/upload", body))
+	if got.err != nil || !bytes.Equal(got.body, chunk) {
+		t.Errorf("app read %d bytes after the first part (%v), want %d",
+			len(got.body), got.err, len(chunk))
+	}
+}
+
+func TestStreamsTheAnswerBody(t *testing.T) {
+	t.Parallel()
+
+	chunk := bytes.Repeat([]byte("y"), chunkSize)
+	firstArrived := make(chan struct{})
+
+	app := startApp(t, func(w http.ResponseWriter, _ *http.Request) {
+		_, _ = w.Write(chunk)
+		_ = http.NewResponseController(w).Flush()
+
+		select {
+		case <-firstArrived:
+			_, _ = w.Write(chunk)
+		case <-time.After(waitLimit):
+		}
+	})
+	addr, _ := startProxy(t, app.URL, nil)
+
+	req := newRequest(t, http.MethodGet, addr, "/download", http.NoBody)
+
+	res, err := newClient(t).Do(req)
+	if err != nil {
+		t.Fatalf("request: %v", err)
+	}
+
+	first := make([]byte, len(chunk))
+	_, err = io.ReadFull(res.Body, first)
+
+	close(firstArrived)
+
+	got := readAnswer(res)
+	if err != nil || got.err != nil || !bytes.Equal(got.body, chunk) {
+		t.Errorf("client read %d bytes after the first part (%v, %v), want %d",
+			len(got.body), err, got.err, len(chunk))
+	}
+}
+
+func TestUpgradedConnectionOutlastsTheTimeouts(t *testing.T) {
+	t.Parallel()
+
+	app := startApp(t, echoAfterUpgrade)
+	addr, out := startProxy(t, app.URL, map[string]string{
+		clientRequestTimeout:    shortTimeoutSetting,
+		clientResponseTimeout:   shortTimeoutSetting,
+		upstreamRequestTimeout:  shortTimeoutSetting,
+		upstreamResponseTimeout: shortTimeoutSetting,
+	})
+
+	conn := dial(t, addr)
+	send(t, conn, "GET /socket HTTP/1.1\r\nHost: app\r\n"+
+		"Connection: Upgrade\r\nUpgrade: websocket\r\n\r\n")
+
+	reader := bufio.NewReader(conn)
+
+	res, err := http.ReadResponse(reader, nil)
+	if err != nil {
+		t.Fatalf("read the answer to the upgrade: %v", err)
+	}
+
+	_ = res.Body.Close()
+
+	if res.StatusCode != http.StatusSwitchingProtocols {
+		t.Fatalf("status %d, want %d", res.StatusCode, http.StatusSwitchingProtocols)
+	}
+
+	// Wait past every timeout, then use the connection.
+	time.Sleep(3 * shortTimeout)
+	send(t, conn, "still here\n")
+
+	echoed, err := reader.ReadString('\n')
+	if err != nil || echoed != "still here\n" {
+		t.Errorf("echo %q (%v), want %q", echoed, err, "still here\n")
+	}
+
+	_ = conn.Close()
+
+	wantLine(t, out.requestLine(t), http.StatusSwitchingProtocols,
+		requestlog.ActionForward)
+}
+
+// echoAfterUpgrade is an app that switches protocols on request, and then
+// sends back each line it receives.
+func echoAfterUpgrade(w http.ResponseWriter, r *http.Request) {
+	if r.Header.Get("Upgrade") != "websocket" {
+		http.Error(w, "not an upgrade", http.StatusBadRequest)
+
+		return
+	}
+
+	conn, buffered, err := http.NewResponseController(w).Hijack()
+	if err != nil {
+		return
+	}
+
+	defer func() {
+		_ = conn.Close()
+	}()
+
+	_, _ = buffered.WriteString("HTTP/1.1 101 Switching Protocols\r\n" +
+		"Connection: Upgrade\r\nUpgrade: websocket\r\n\r\n")
+	_ = buffered.Flush()
+
+	for {
+		line, err := buffered.ReadString('\n')
+		if err != nil {
+			return
+		}
+
+		_, _ = buffered.WriteString(line)
+		_ = buffered.Flush()
+	}
+}
+
+func TestServerHasTheFixedLimits(t *testing.T) {
+	t.Parallel()
+
+	cfg, err := config.FromEnvironment(func(string) (string, bool) { return "", false })
+	if err != nil {
+		t.Fatalf("default settings: %v", err)
+	}
+
+	server := proxy.New(proxy.Params{
+		Config:     cfg,
+		RequestLog: io.Discard,
+		ProcessLog: requestlog.NewProcessLogger(io.Discard),
+	})
+
+	if server.Addr != ":8080" || server.MaxHeaderBytes != 28<<10 ||
+		server.IdleTimeout != 2*time.Minute || server.ReadHeaderTimeout != time.Minute {
+		t.Errorf("server listens on %q with header limit %d, idle time %s and "+
+			"header timeout %s", server.Addr, server.MaxHeaderBytes,
+			server.IdleTimeout, server.ReadHeaderTimeout)
+	}
+}
+
+func TestRefusesHeadersOver32KiB(t *testing.T) {
+	t.Parallel()
+
+	var calls atomic.Int32
+
+	app := startApp(t, func(http.ResponseWriter, *http.Request) {
+		calls.Add(1)
+	})
+	addr, _ := startProxy(t, app.URL, nil)
+
+	// size counts every byte of the request: the request line, the
+	// headers and the blank line that ends them.
+	const (
+		start = "GET / HTTP/1.1\r\nHost: app\r\nX-Large: "
+		end   = "\r\n\r\n"
+	)
+
+	for _, tc := range []struct {
+		size int
+		want int
+	}{
+		{size: 32 << 10, want: http.StatusOK},
+		{size: 32<<10 + 1, want: http.StatusRequestHeaderFieldsTooLarge},
+	} {
+		conn := dial(t, addr)
+		send(t, conn, start+strings.Repeat("a", tc.size-len(start)-len(end))+end)
+		wantStatus(t, readResponse(t, conn), tc.want)
+	}
+
+	if calls.Load() != 1 {
+		t.Errorf("the app was called %d times, want once", calls.Load())
+	}
+}
+
+func TestAnswers502WhenTheAppCannotBeReached(t *testing.T) {
+	t.Parallel()
+
+	listener, err := (&net.ListenConfig{}).Listen(t.Context(), "tcp", localhost+":0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+
+	closedAddr := listener.Addr().String()
+	_ = listener.Close()
+
+	addr, out := startProxy(t, "http://"+closedAddr, nil)
+
+	wantStatus(t, get(t, addr, "/"), http.StatusBadGateway)
+	wantLine(t, out.requestLine(t), http.StatusBadGateway,
+		requestlog.ActionUpstreamError)
+
+	logged := slices.ContainsFunc(out.lines(t), func(line map[string]any) bool {
+		return line["type"] == "process" && line["msg"] == "request to the app failed"
+	})
+	if !logged {
+		t.Errorf("no process line says the request to the app failed")
+	}
+}
+
+func TestLogsAnAnswerThatBrokeOff(t *testing.T) {
+	t.Parallel()
+
+	app := startApp(t, func(w http.ResponseWriter, _ *http.Request) {
+		_, _ = io.WriteString(w, "the first part")
+		_ = http.NewResponseController(w).Flush()
+
+		panic(http.ErrAbortHandler) // drops the connection mid-answer
+	})
+	addr, out := startProxy(t, app.URL, nil)
+
+	got := get(t, addr, "/")
+	if string(got.body) != "the first part" || !errors.Is(got.err, io.ErrUnexpectedEOF) {
+		t.Errorf("client read %q (%v), want the first part cut off", got.body, got.err)
+	}
+
+	wantLine(t, out.requestLine(t), http.StatusOK, requestlog.ActionUpstreamError)
+}
+
+func TestLogsAClientThatWentAway(t *testing.T) {
+	t.Parallel()
+
+	arrived := make(chan struct{})
+
+	app := startApp(t, func(_ http.ResponseWriter, r *http.Request) {
+		close(arrived)
+		<-r.Context().Done()
+	})
+	addr, out := startProxy(t, app.URL, nil)
+
+	conn := dial(t, addr)
+	send(t, conn, "GET /slow HTTP/1.1\r\nHost: app\r\n\r\n")
+
+	select {
+	case <-arrived:
+	case <-time.After(waitLimit):
+		t.Fatal("the request never reached the app")
+	}
+
+	_ = conn.Close()
+
+	line := out.requestLine(t)
+	if !line.Aborted || line.Status != 0 || line.Action != requestlog.ActionForward {
+		t.Errorf("log line has aborted %v, status %d and action %q, "+
+			"want true, 0 and %q",
+			line.Aborted, line.Status, line.Action, requestlog.ActionForward)
+	}
+}

internal/proxy/proxy.go

@@ -0,0 +1,105 @@
+// Package proxy passes each request to the app and the app's answer back,
+// unchanged, within the size and time limits, and writes one request log
+// line for each request.
+package proxy
+
+import (
+	"io"
+	"log"
+	"log/slog"
+	"net/http"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/config"
+)
+
+// The request line and headers a client may send, and how long a
+// kept-open client connection may wait for its next request, are fixed
+// rather than settings. The limit on the request line and headers is
+// 32 KiB, but Go's server reads 4 KiB past its MaxHeaderBytes before it
+// refuses, so MaxHeaderBytes is set 4 KiB lower. The idle time is longer
+// than the 90 seconds after which traefik closes a connection it is not
+// using, so traefik never sends a request on a connection smallwebwaf is
+// closing.
+const (
+	requestHeaderMaxBytes = 32<<10 - 4<<10
+	clientIdleTimeout     = 120 * time.Second
+)
+
+// How smallwebwaf keeps connections to the app open between requests.
+const (
+	appIdleConns       = 100
+	appIdleConnTimeout = 90 * time.Second
+)
+
+// Params are what New needs.
+type Params struct {
+	Config *config.Config
+	// RequestLog receives one JSON line per request.
+	RequestLog io.Writer
+	// ProcessLog receives the process's own messages.
+	ProcessLog *slog.Logger
+}
+
+// New returns the server smallwebwaf runs: each request it reads passes
+// through the proxy. Go's server itself refuses headers over 32 KiB, with
+// 431, closes a connection idle for 120 seconds, and applies
+// SWWAF_CLIENT_REQUEST_TIMEOUT while the headers arrive; the proxy
+// applies the timeouts and size limits from then on.
+func New(params Params) *http.Server {
+	errorLog := slog.NewLogLogger(params.ProcessLog.Handler(), slog.LevelWarn)
+
+	return &http.Server{
+		Addr: params.Config.ListenAddr,
+		Handler: &handler{
+			config:     params.Config,
+			requestLog: params.RequestLog,
+			processLog: params.ProcessLog,
+			errorLog:   errorLog,
+			transport:  newTransport(),
+		},
+		ReadHeaderTimeout: params.Config.ClientRequestTimeout,
+		IdleTimeout:       clientIdleTimeout,
+		MaxHeaderBytes:    requestHeaderMaxBytes,
+		ErrorLog:          errorLog,
+	}
+}
+
+// handler is the proxy. It holds what every request shares; what belongs
+// to one request is in a request.
+type handler struct {
+	config     *config.Config
+	requestLog io.Writer
+	processLog *slog.Logger
+	errorLog   *log.Logger
+	transport  http.RoundTripper
+}
+
+// newTransport returns what carries requests to the app. It never goes
+// through a proxy named in the environment, and leaves the app's answers
+// compressed or not as the app sent them.
+func newTransport() *http.Transport {
+	return &http.Transport{
+		MaxIdleConns:        appIdleConns,
+		MaxIdleConnsPerHost: appIdleConns,
+		IdleConnTimeout:     appIdleConnTimeout,
+		DisableCompression:  true,
+	}
+}
+
+// ServeHTTP handles one request: it works out the client, runs the
+// checks, passes the request to the app and the answer back within the
+// limits, and writes the request's log line.
+func (h *handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	rq := h.newRequest(w, r)
+	defer rq.finish()
+
+	refused := rq.check()
+	if refused != nil {
+		rq.answer(*refused)
+
+		return
+	}
+
+	rq.forward(r.Context())
+}

internal/proxy/proxy_test.go

@@ -0,0 +1,328 @@
+package proxy_test
+
+import (
+	"bufio"
+	"bytes"
+	"encoding/json"
+	"io"
+	"maps"
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/config"
+	"sneak.berlin/go/smallwebwaf/internal/proxy"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+const (
+	// shortTimeout is what a test sets a timeout to, to see it run out.
+	shortTimeout = 300 * time.Millisecond
+	// shortTimeoutSetting is shortTimeout as a setting's value.
+	shortTimeoutSetting = "300ms"
+	// longTimeoutSetting is a timeout that does not run out in a test.
+	longTimeoutSetting = "10s"
+	// waitLimit bounds how long a test waits for what should happen.
+	waitLimit = 10 * time.Second
+	// pollInterval is how often a test looks for a log line.
+	pollInterval = 10 * time.Millisecond
+	// localhost is where every test server listens, and so the address
+	// smallwebwaf sees each test's requests come from.
+	localhost = "127.0.0.1"
+)
+
+// The settings the tests set.
+const (
+	clientRequestTimeout    = "SWWAF_CLIENT_REQUEST_TIMEOUT"
+	clientResponseTimeout   = "SWWAF_CLIENT_RESPONSE_TIMEOUT"
+	upstreamRequestTimeout  = "SWWAF_UPSTREAM_REQUEST_TIMEOUT"
+	upstreamResponseTimeout = "SWWAF_UPSTREAM_RESPONSE_TIMEOUT"
+	requestMaxBytes         = "SWWAF_REQUEST_MAX_BYTES"
+	responseMaxBytes        = "SWWAF_RESPONSE_MAX_BYTES"
+	trustedProxies          = "SWWAF_TRUSTED_PROXIES"
+)
+
+// output collects what smallwebwaf writes on stdout.
+type output struct {
+	mu  sync.Mutex
+	buf bytes.Buffer
+}
+
+// Write adds lines smallwebwaf writes.
+func (o *output) Write(p []byte) (int, error) {
+	o.mu.Lock()
+	defer o.mu.Unlock()
+
+	return o.buf.Write(p)
+}
+
+// lines returns every line written so far, decoded.
+func (o *output) lines(t *testing.T) []map[string]any {
+	t.Helper()
+	o.mu.Lock()
+	defer o.mu.Unlock()
+
+	var lines []map[string]any
+
+	for text := range strings.Lines(o.buf.String()) {
+		var line map[string]any
+
+		err := json.Unmarshal([]byte(text), &line)
+		if err != nil {
+			t.Fatalf("output line %q is not JSON: %v", text, err)
+		}
+
+		lines = append(lines, line)
+	}
+
+	return lines
+}
+
+// logLine is a request log line, as typed fields and as the JSON object
+// it was written as.
+type logLine struct {
+	requestlog.Line
+
+	fields map[string]any
+}
+
+// requestLines waits for count request log lines and returns them.
+func (o *output) requestLines(t *testing.T, count int) []logLine {
+	t.Helper()
+
+	deadline := time.Now().Add(waitLimit)
+	for time.Now().Before(deadline) {
+		var found []logLine
+
+		for _, fields := range o.lines(t) {
+			if fields["type"] == "request" {
+				found = append(found, decodeLine(t, fields))
+			}
+		}
+
+		if len(found) >= count {
+			return found
+		}
+
+		time.Sleep(pollInterval)
+	}
+
+	t.Fatalf("fewer than %d request log lines after %s", count, waitLimit)
+
+	return nil
+}
+
+// requestLine waits for the request log line of a test's one request.
+func (o *output) requestLine(t *testing.T) logLine {
+	t.Helper()
+
+	return o.requestLines(t, 1)[0]
+}
+
+// decodeLine reads a request log line's fields into a logLine.
+func decodeLine(t *testing.T, fields map[string]any) logLine {
+	t.Helper()
+
+	encoded, err := json.Marshal(fields)
+	if err != nil {
+		t.Fatalf("encode %v: %v", fields, err)
+	}
+
+	line := logLine{fields: fields}
+
+	err = json.Unmarshal(encoded, &line.Line)
+	if err != nil {
+		t.Fatalf("decode %s: %v", encoded, err)
+	}
+
+	return line
+}
+
+// startApp starts app as the app smallwebwaf passes requests to.
+func startApp(t *testing.T, app http.HandlerFunc) *httptest.Server {
+	t.Helper()
+
+	server := httptest.NewServer(app)
+	t.Cleanup(server.Close)
+
+	return server
+}
+
+// startProxy starts smallwebwaf in front of the app at appURL, with the
+// settings in env on top of the defaults, and returns where it listens and
+// what it writes.
+func startProxy(t *testing.T, appURL string, env map[string]string) (string, *output) {
+	t.Helper()
+
+	settings := map[string]string{"SWWAF_UPSTREAM_URL": appURL}
+	maps.Copy(settings, env)
+
+	cfg, err := config.FromEnvironment(func(name string) (string, bool) {
+		value, ok := settings[name]
+
+		return value, ok
+	})
+	if err != nil {
+		t.Fatalf("settings %v: %v", settings, err)
+	}
+
+	out := &output{}
+	server := proxy.New(proxy.Params{
+		Config:     cfg,
+		RequestLog: out,
+		ProcessLog: requestlog.NewProcessLogger(out),
+	})
+
+	listener, err := (&net.ListenConfig{}).Listen(t.Context(), "tcp", localhost+":0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+
+	go func() {
+		_ = server.Serve(listener)
+	}()
+
+	t.Cleanup(func() {
+		_ = server.Close()
+	})
+
+	return listener.Addr().String(), out
+}
+
+// newClient returns an HTTP client that sends requests as they are made,
+// with no compression of its own.
+func newClient(t *testing.T) *http.Client {
+	t.Helper()
+
+	transport := &http.Transport{DisableCompression: true}
+	t.Cleanup(transport.CloseIdleConnections)
+
+	return &http.Client{Transport: transport}
+}
+
+// answer is a response as a test reads it: the status, the headers, as
+// much of the body as arrived, and the error that ended the reading, nil
+// when the whole body arrived.
+type answer struct {
+	status int
+	header http.Header
+	body   []byte
+	err    error
+}
+
+// readAnswer reads all of res, and closes its body.
+func readAnswer(res *http.Response) answer {
+	body, err := io.ReadAll(res.Body)
+	_ = res.Body.Close()
+
+	return answer{status: res.StatusCode, header: res.Header, body: body, err: err}
+}
+
+// newRequest makes a request for path to smallwebwaf at addr.
+func newRequest(t *testing.T, method, addr, path string, body io.Reader) *http.Request {
+	t.Helper()
+
+	req, err := http.NewRequestWithContext(t.Context(), method, "http://"+addr+path, body)
+	if err != nil {
+		t.Fatalf("new request: %v", err)
+	}
+
+	return req
+}
+
+// do sends req and reads the answer.
+func do(t *testing.T, req *http.Request) answer {
+	t.Helper()
+
+	res, err := newClient(t).Do(req)
+	if err != nil {
+		t.Fatalf("%s %s: %v", req.Method, req.URL.Path, err)
+	}
+
+	return readAnswer(res)
+}
+
+// get sends a GET request for path to smallwebwaf at addr.
+func get(t *testing.T, addr, path string) answer {
+	t.Helper()
+
+	return do(t, newRequest(t, http.MethodGet, addr, path, http.NoBody))
+}
+
+// dial opens a connection to smallwebwaf at addr, for requests the HTTP
+// client cannot make, such as one that stops sending halfway.
+func dial(t *testing.T, addr string) net.Conn {
+	t.Helper()
+
+	conn, err := (&net.Dialer{}).DialContext(t.Context(), "tcp", addr)
+	if err != nil {
+		t.Fatalf("dial %s: %v", addr, err)
+	}
+
+	t.Cleanup(func() {
+		_ = conn.Close()
+	})
+
+	return conn
+}
+
+// send writes text to conn.
+func send(t *testing.T, conn net.Conn, text string) {
+	t.Helper()
+
+	_, err := io.WriteString(conn, text)
+	if err != nil {
+		t.Fatalf("send: %v", err)
+	}
+}
+
+// readResponse reads the answer to a request sent on conn.
+func readResponse(t *testing.T, conn net.Conn) answer {
+	t.Helper()
+
+	err := conn.SetReadDeadline(time.Now().Add(waitLimit))
+	if err != nil {
+		t.Fatalf("set read deadline: %v", err)
+	}
+
+	res, err := http.ReadResponse(bufio.NewReader(conn), nil)
+	if err != nil {
+		t.Fatalf("read response: %v", err)
+	}
+
+	return readAnswer(res)
+}
+
+// wantLine checks the request log line's status and action.
+func wantLine(t *testing.T, line logLine, status int, action string) {
+	t.Helper()
+
+	if line.Status != status || line.Action != action {
+		t.Errorf("log line has status %d and action %q, want %d and %q",
+			line.Status, line.Action, status, action)
+	}
+}
+
+// wantStatus checks an answer's status.
+func wantStatus(t *testing.T, got answer, status int) {
+	t.Helper()
+
+	if got.status != status {
+		t.Errorf("status %d, want %d", got.status, status)
+	}
+}
+
+// wantTimedOut checks that what began at start ended once shortTimeout
+// had run out, and not much later.
+func wantTimedOut(t *testing.T, start time.Time) {
+	t.Helper()
+
+	took := time.Since(start)
+	if took < shortTimeout || took > shortTimeout+waitLimit/2 {
+		t.Errorf("took %s, want %s", took, shortTimeout)
+	}
+}

internal/proxy/request.go

@@ -0,0 +1,452 @@
+package proxy
+
+import (
+	"context"
+	"errors"
+	"net/http"
+	"net/http/httptrace"
+	"net/http/httputil"
+	"net/netip"
+	"os"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// flushAfterEachWrite has ReverseProxy pass on each part of the app's
+// answer as soon as it arrives.
+const flushAfterEachWrite time.Duration = -1
+
+// refusal is smallwebwaf refusing a request, or refusing to go on with it:
+// the status the client is answered if the response has not started yet,
+// and the action the log line names.
+type refusal struct {
+	status int
+	action string
+}
+
+// request is one request on its way through smallwebwaf, from the moment
+// its headers have been read to its log line.
+type request struct {
+	h  *handler
+	in *http.Request
+	// rc sets the deadlines of the connection to the client.
+	rc   *http.ResponseController
+	out  *responseWriter
+	body *requestBody // nil for a request without a body
+	line requestlog.Line
+
+	peer        netip.Addr
+	peerTrusted bool
+	start       time.Time
+	// upstreamStart is when the request was handed to the app.
+	upstreamStart time.Time
+	// cancel ends the request to the app.
+	cancel context.CancelFunc
+	// refused is the first refusal, from whichever goroutine meets it.
+	refused atomic.Pointer[refusal]
+	// complete is true once the app's whole answer has been passed on.
+	complete bool
+
+	// mu guards what follows. The timeouts run on goroutines of their
+	// own, and the transport starts and stops them from its own; once
+	// timersStopped is set, none of them acts any more.
+	mu                    sync.Mutex
+	timersStopped         bool
+	clientRequestTimer    *time.Timer
+	upstreamRequestTimer  *time.Timer
+	upstreamResponseTimer *time.Timer
+	// requestSent is when the app had been sent the whole request.
+	requestSent time.Time
+}
+
+// newRequest starts handling r: it notes the time and works out the
+// client.
+func (h *handler) newRequest(w http.ResponseWriter, r *http.Request) *request {
+	start := time.Now()
+	peer := peerAddress(r)
+	trusted := h.config.TrustedProxies
+	client := clientAddress(peer, r.Header.Values("X-Forwarded-For"), trusted)
+
+	rq := &request{
+		h:           h,
+		in:          r,
+		rc:          http.NewResponseController(w),
+		out:         &responseWriter{ResponseWriter: w},
+		peer:        peer,
+		peerTrusted: isInside(peer, trusted),
+		start:       start,
+		line: requestlog.Line{
+			Time:      requestlog.FormatTime(start),
+			ClientIP:  client.String(),
+			PeerIP:    peer.String(),
+			Method:    r.Method,
+			Host:      r.Host,
+			Path:      r.URL.EscapedPath(),
+			Query:     r.URL.RawQuery,
+			Protocol:  r.Proto,
+			Referer:   r.Referer(),
+			UserAgent: r.UserAgent(),
+			Action:    requestlog.ActionForward,
+		},
+	}
+	if r.Body != http.NoBody {
+		rq.body = &requestBody{body: limitBody(r.Body, h.config.RequestMaxBytes), rq: rq}
+	}
+
+	return rq
+}
+
+// check is the one place where a request can be refused once its client
+// is known, before its body is read or anything reaches the app; the rate
+// limits and country lists of milestone 2 go here. It returns nil to let
+// the request through.
+func (rq *request) check() *refusal {
+	maxBytes := rq.h.config.RequestMaxBytes
+	if maxBytes > 0 && rq.in.ContentLength > maxBytes {
+		return &refusal{
+			status: http.StatusRequestEntityTooLarge,
+			action: requestlog.ActionTooLarge,
+		}
+	}
+
+	return nil
+}
+
+// forward passes the request to the app and the app's answer back. ctx
+// is the request's own context.
+func (rq *request) forward(ctx context.Context) {
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	rq.cancel = cancel
+	ctx = httptrace.WithClientTrace(ctx, &httptrace.ClientTrace{
+		WroteRequest: rq.wroteRequest,
+	})
+
+	out := rq.in.WithContext(ctx)
+	if rq.body != nil {
+		out.Body = rq.body
+	}
+
+	reverseProxy := &httputil.ReverseProxy{
+		Rewrite:        rq.rewrite,
+		Transport:      rq.h.transport,
+		FlushInterval:  flushAfterEachWrite,
+		ErrorLog:       rq.h.errorLog,
+		ModifyResponse: rq.modifyResponse,
+		ErrorHandler:   rq.answerError,
+	}
+
+	rq.startRequestTimers()
+	rq.upstreamStart = time.Now()
+	reverseProxy.ServeHTTP(rq.out, out)
+}
+
+// rewrite makes the request the app receives: the client's request,
+// unchanged, sent to SWWAF_UPSTREAM_URL, with the forwarded headers set.
+func (rq *request) rewrite(pr *httputil.ProxyRequest) {
+	upstream := rq.h.config.UpstreamURL
+	pr.Out.URL.Scheme = upstream.Scheme
+	pr.Out.URL.Host = upstream.Host
+	// ReverseProxy drops query parameters it cannot parse; the app gets
+	// the query as the client sent it.
+	pr.Out.URL.RawQuery = pr.In.URL.RawQuery
+	setForwardedHeaders(pr.In, pr.Out, rq.peer, rq.peerTrusted)
+}
+
+// modifyResponse looks at the app's answer before ReverseProxy passes it
+// on.
+func (rq *request) modifyResponse(res *http.Response) error {
+	rq.line.UpstreamStatus = res.StatusCode
+
+	if res.StatusCode == http.StatusSwitchingProtocols {
+		// An upgraded connection, such as a WebSocket, is not cut by the
+		// timeouts. ReverseProxy writes this answer straight to the
+		// connection it takes over, not through rq.out.
+		rq.stopTimers()
+		rq.out.status = res.StatusCode
+
+		return nil
+	}
+
+	maxBytes := rq.h.config.ResponseMaxBytes
+	if maxBytes > 0 && res.Body != http.NoBody && res.ContentLength > maxBytes {
+		rq.refuse(refusal{status: http.StatusBadGateway, action: requestlog.ActionTooLarge})
+
+		return errResponseTooLarge
+	}
+
+	res.Body = &responseBody{body: limitBody(res.Body, maxBytes), rq: rq}
+	rq.startClientResponseTimeout()
+
+	return nil
+}
+
+// answerError is ReverseProxy's ErrorHandler: the request could not be
+// passed to the app, or the app's answer cannot be passed on.
+func (rq *request) answerError(_ http.ResponseWriter, _ *http.Request, err error) {
+	refused := rq.refused.Load()
+	if refused == nil {
+		if rq.in.Context().Err() != nil {
+			return // the client has gone, and there is no one to answer
+		}
+
+		rq.h.processLog.Warn("request to the app failed", "error", err.Error())
+
+		refused = &refusal{
+			status: http.StatusBadGateway,
+			action: requestlog.ActionUpstreamError,
+		}
+	}
+
+	rq.answer(*refused)
+}
+
+// answer sends smallwebwaf's own answer, unless the response has already
+// started, and records the refusal for the log line.
+func (rq *request) answer(r refusal) {
+	rq.refused.CompareAndSwap(nil, &r)
+
+	if rq.out.status != 0 {
+		return // too late to answer: the connection can only be cut
+	}
+
+	// A client found too slow is read no more; any other may go on
+	// sending until its time is up, so that Go's server can read the
+	// rest of the body and end the request cleanly.
+	deadline := rq.clientRequestDeadline()
+	if r.status == http.StatusRequestTimeout {
+		deadline = time.Now()
+	}
+
+	rq.stopReadingBody(deadline)
+
+	timeout := rq.h.config.ClientResponseTimeout
+	if timeout > 0 {
+		_ = rq.rc.SetWriteDeadline(time.Now().Add(timeout))
+	}
+
+	http.Error(rq.out, http.StatusText(r.status), r.status)
+}
+
+// refuse records r, unless an earlier refusal was, and ends the request
+// to the app.
+func (rq *request) refuse(r refusal) {
+	rq.refused.CompareAndSwap(nil, &r)
+	rq.cancel()
+}
+
+// finish ends the request's timeouts and writes its log line.
+func (rq *request) finish() {
+	rq.stopTimers()
+
+	refused := rq.refused.Load()
+	if refused == nil {
+		rq.stopReadingBody(rq.clientRequestDeadline())
+	}
+
+	line := &rq.line
+	line.Status = rq.out.status
+	line.ResponseBytes = rq.out.bytes
+
+	if rq.body != nil {
+		line.RequestBytes = rq.body.bytes.Load()
+	}
+
+	switch {
+	case refused != nil:
+		line.Action = refused.action
+	case errors.Is(rq.out.err, os.ErrDeadlineExceeded):
+		// The client took longer than SWWAF_CLIENT_RESPONSE_TIMEOUT to
+		// take the response.
+		line.Action = requestlog.ActionTimedOut
+	case !rq.complete && (rq.out.err != nil || rq.in.Context().Err() != nil):
+		line.Aborted = true
+	}
+
+	now := time.Now()
+	line.DurationTotal = requestlog.Milliseconds(now.Sub(rq.start))
+
+	if !rq.upstreamStart.IsZero() {
+		line.DurationUpstreamTotal = requestlog.Milliseconds(now.Sub(rq.upstreamStart))
+	}
+
+	err := requestlog.Write(rq.h.requestLog, line)
+	if err != nil {
+		rq.h.processLog.Error("writing the request log failed", "error", err.Error())
+	}
+}
+
+// clientRequestDeadline is when the client must have sent its whole
+// request, or zero when SWWAF_CLIENT_REQUEST_TIMEOUT is off.
+func (rq *request) clientRequestDeadline() time.Time {
+	timeout := rq.h.config.ClientRequestTimeout
+	if timeout == 0 {
+		return time.Time{}
+	}
+
+	return rq.start.Add(timeout)
+}
+
+// stopReadingBody ends, at deadline, the reading of a client body that has
+// not arrived whole: Go's server then reads no more of it, and closes the
+// connection after the answer.
+func (rq *request) stopReadingBody(deadline time.Time) {
+	if rq.body == nil || rq.body.received.Load() {
+		return
+	}
+
+	_ = rq.rc.SetReadDeadline(deadline)
+}
+
+// startRequestTimers starts the timeouts that run while the request goes
+// to the app: SWWAF_CLIENT_REQUEST_TIMEOUT until the client has sent its
+// whole body, and SWWAF_UPSTREAM_REQUEST_TIMEOUT until the app has been
+// sent the whole request.
+func (rq *request) startRequestTimers() {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	if rq.body != nil && rq.h.config.ClientRequestTimeout > 0 {
+		rq.clientRequestTimer = time.AfterFunc(
+			time.Until(rq.clientRequestDeadline()), rq.requestTimedOut)
+	}
+
+	timeout := rq.h.config.UpstreamRequestTimeout
+	if timeout > 0 {
+		rq.upstreamRequestTimer = time.AfterFunc(timeout, rq.requestTimedOut)
+	}
+}
+
+// requestTimedOut is called when a request timeout runs out while the
+// request is still on its way to the app. The answer names the side
+// smallwebwaf was waiting on at that moment: 408 when it was waiting for
+// the client to send more of its body, 504 when it was waiting for the
+// app to be reached or to take what it had.
+func (rq *request) requestTimedOut() {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	if rq.timersStopped {
+		return
+	}
+
+	if rq.body == nil || !rq.body.waiting.Load() {
+		rq.refuse(refusal{
+			status: http.StatusGatewayTimeout,
+			action: requestlog.ActionTimedOut,
+		})
+
+		return
+	}
+
+	rq.refuse(refusal{
+		status: http.StatusRequestTimeout,
+		action: requestlog.ActionTimedOut,
+	})
+	// The transport gives up on the app only once its Read of the
+	// client's body returns, so that Read is ended now. The lock keeps
+	// this from reaching the connection after the request is handled.
+	_ = rq.rc.SetReadDeadline(time.Now())
+}
+
+// bodyReceived is called once the client has sent its whole body.
+func (rq *request) bodyReceived() {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	stopTimer(rq.clientRequestTimer)
+}
+
+// wroteRequest is called once the app has been sent the whole request:
+// the request timeouts end and SWWAF_UPSTREAM_RESPONSE_TIMEOUT starts.
+func (rq *request) wroteRequest(info httptrace.WroteRequestInfo) {
+	if info.Err != nil {
+		return // the transport gives up, or tries again
+	}
+
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	if rq.timersStopped {
+		return
+	}
+
+	stopTimer(rq.clientRequestTimer)
+	stopTimer(rq.upstreamRequestTimer)
+	rq.requestSent = time.Now()
+
+	timeout := rq.h.config.UpstreamResponseTimeout
+	if timeout > 0 {
+		rq.upstreamResponseTimer = time.AfterFunc(timeout, rq.responseTimedOut)
+	}
+}
+
+// responseTimedOut is called when SWWAF_UPSTREAM_RESPONSE_TIMEOUT runs out
+// before the app has sent its whole answer.
+func (rq *request) responseTimedOut() {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	if !rq.timersStopped {
+		rq.refuse(refusal{
+			status: http.StatusGatewayTimeout,
+			action: requestlog.ActionTimedOut,
+		})
+	}
+}
+
+// responseReceived is called once the app has sent its whole answer.
+func (rq *request) responseReceived() {
+	rq.complete = true
+	rq.stopTimers()
+}
+
+// startClientResponseTimeout sets SWWAF_CLIENT_RESPONSE_TIMEOUT on the
+// connection to the client: the response must reach the client within it
+// of the end of the request, or of now if the app answers before it has
+// the whole request.
+func (rq *request) startClientResponseTimeout() {
+	timeout := rq.h.config.ClientResponseTimeout
+	if timeout == 0 {
+		return
+	}
+
+	from := rq.sentAt()
+	if from.IsZero() {
+		from = time.Now()
+	}
+
+	_ = rq.rc.SetWriteDeadline(from.Add(timeout))
+}
+
+// sentAt is when the app had been sent the whole request, or zero.
+func (rq *request) sentAt() time.Time {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	return rq.requestSent
+}
+
+// stopTimers stops the request's timeouts and keeps any from starting
+// later: the app's answer is complete, the connection upgraded, or the
+// request handled.
+func (rq *request) stopTimers() {
+	rq.mu.Lock()
+	defer rq.mu.Unlock()
+
+	rq.timersStopped = true
+	stopTimer(rq.clientRequestTimer)
+	stopTimer(rq.upstreamRequestTimer)
+	stopTimer(rq.upstreamResponseTimer)
+}
+
+// stopTimer stops t, which is nil when its timeout is off.
+func stopTimer(t *time.Timer) {
+	if t != nil {
+		t.Stop()
+	}
+}

internal/proxy/timeouts_test.go

@@ -0,0 +1,258 @@
+package proxy_test
+
+import (
+	"errors"
+	"io"
+	"net"
+	"net/http"
+	"strconv"
+	"sync"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// largeBodySize is more than the connections between the client,
+// smallwebwaf and the app can hold while nobody reads, so that a sender
+// soon waits.
+const largeBodySize = 64 << 20
+
+// writeSize is how much a test sender writes at a time.
+const writeSize = 32 << 10
+
+func TestRequestTimeouts(t *testing.T) {
+	t.Parallel()
+
+	for _, tc := range []struct {
+		name string
+		env  map[string]string
+		// appTakesNothing has the app never read, while the client sends
+		// as fast as it can; otherwise the app reads, and the client
+		// stops sending halfway.
+		appTakesNothing bool
+		want            int
+	}{
+		{
+			name: "client request timeout, waiting on the client",
+			env:  map[string]string{clientRequestTimeout: shortTimeoutSetting},
+			want: http.StatusRequestTimeout,
+		},
+		{
+			name: "upstream request timeout, waiting on the client",
+			env: map[string]string{
+				upstreamRequestTimeout: shortTimeoutSetting,
+				clientRequestTimeout:   longTimeoutSetting,
+			},
+			want: http.StatusRequestTimeout,
+		},
+		{
+			name:            "upstream request timeout, waiting on the app",
+			env:             map[string]string{upstreamRequestTimeout: shortTimeoutSetting},
+			appTakesNothing: true,
+			want:            http.StatusGatewayTimeout,
+		},
+		{
+			name: "client request timeout, waiting on the app",
+			env: map[string]string{
+				clientRequestTimeout:   shortTimeoutSetting,
+				upstreamRequestTimeout: longTimeoutSetting,
+			},
+			appTakesNothing: true,
+			want:            http.StatusGatewayTimeout,
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			var (
+				appURL      string
+				sendRequest func(*testing.T, string) net.Conn
+			)
+
+			if tc.appTakesNothing {
+				appURL, sendRequest = startAppThatTakesNothing(t), sendLargeBody
+			} else {
+				appURL, sendRequest = startApp(t, readBody).URL, sendPartOfBody
+			}
+
+			addr, out := startProxy(t, appURL, tc.env)
+			start := time.Now()
+			conn := sendRequest(t, addr)
+
+			wantStatus(t, readResponse(t, conn), tc.want)
+			wantTimedOut(t, start)
+			wantLine(t, out.requestLine(t), tc.want, requestlog.ActionTimedOut)
+		})
+	}
+}
+
+// readBody is an app that reads the request body, then answers.
+func readBody(_ http.ResponseWriter, r *http.Request) {
+	_, _ = io.Copy(io.Discard, r.Body)
+}
+
+// startAppThatTakesNothing starts an app that accepts connections and
+// never reads from them, and returns its URL.
+func startAppThatTakesNothing(t *testing.T) string {
+	t.Helper()
+
+	listener, err := (&net.ListenConfig{}).Listen(t.Context(), "tcp", localhost+":0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+
+	var (
+		mu   sync.Mutex
+		held []net.Conn
+	)
+
+	hold := func(conn net.Conn) {
+		mu.Lock()
+		defer mu.Unlock()
+
+		held = append(held, conn)
+	}
+
+	go func() {
+		for {
+			conn, err := listener.Accept()
+			if err != nil {
+				return
+			}
+
+			hold(conn)
+		}
+	}()
+
+	t.Cleanup(func() {
+		_ = listener.Close()
+
+		mu.Lock()
+		defer mu.Unlock()
+
+		for _, conn := range held {
+			_ = conn.Close()
+		}
+	})
+
+	return "http://" + listener.Addr().String()
+}
+
+// sendPartOfBody sends a request that announces a large body, and only
+// the first bytes of it.
+func sendPartOfBody(t *testing.T, addr string) net.Conn {
+	t.Helper()
+
+	conn := dial(t, addr)
+	send(t, conn, "POST /upload HTTP/1.1\r\nHost: app\r\nContent-Length: "+
+		strconv.Itoa(largeBodySize)+"\r\n\r\nthe first bytes")
+
+	return conn
+}
+
+// sendLargeBody sends a request with a large body, as fast as smallwebwaf
+// takes it, from a goroutine of its own.
+func sendLargeBody(t *testing.T, addr string) net.Conn {
+	t.Helper()
+
+	conn := dial(t, addr)
+	send(t, conn, "POST /upload HTTP/1.1\r\nHost: app\r\nContent-Length: "+
+		strconv.Itoa(largeBodySize)+"\r\n\r\n")
+
+	go func() {
+		chunk := make([]byte, writeSize)
+		for range largeBodySize / writeSize {
+			_, err := conn.Write(chunk)
+			if err != nil {
+				return
+			}
+		}
+	}()
+
+	return conn
+}
+
+func TestAppTooSlowToAnswer(t *testing.T) {
+	t.Parallel()
+
+	app := startApp(t, func(_ http.ResponseWriter, r *http.Request) {
+		<-r.Context().Done()
+	})
+	addr, out := startProxy(t, app.URL, map[string]string{
+		upstreamResponseTimeout: shortTimeoutSetting,
+	})
+
+	start := time.Now()
+
+	wantStatus(t, get(t, addr, "/slow"), http.StatusGatewayTimeout)
+	wantTimedOut(t, start)
+
+	line := out.requestLine(t)
+	wantLine(t, line, http.StatusGatewayTimeout, requestlog.ActionTimedOut)
+
+	_, answered := line.fields["upstream_status"]
+	if answered {
+		t.Errorf("log line has upstream_status %v for an app that never answered",
+			line.fields["upstream_status"])
+	}
+}
+
+func TestAppTooSlowToFinishItsAnswer(t *testing.T) {
+	t.Parallel()
+
+	app := startApp(t, func(w http.ResponseWriter, r *http.Request) {
+		_, _ = io.WriteString(w, "the first part")
+		_ = http.NewResponseController(w).Flush()
+
+		<-r.Context().Done()
+	})
+	addr, out := startProxy(t, app.URL, map[string]string{
+		upstreamResponseTimeout: shortTimeoutSetting,
+	})
+
+	start := time.Now()
+	got := get(t, addr, "/slow")
+	wantStatus(t, got, http.StatusOK)
+
+	if string(got.body) != "the first part" || !errors.Is(got.err, io.ErrUnexpectedEOF) {
+		t.Errorf("client read %q (%v), want the first part cut off", got.body, got.err)
+	}
+
+	wantTimedOut(t, start)
+
+	line := out.requestLine(t)
+	wantLine(t, line, http.StatusOK, requestlog.ActionTimedOut)
+
+	if line.UpstreamStatus != http.StatusOK {
+		t.Errorf("log line has upstream_status %d, want %d",
+			line.UpstreamStatus, http.StatusOK)
+	}
+}
+
+func TestClientTooSlowToTakeTheAnswer(t *testing.T) {
+	t.Parallel()
+
+	app := startApp(t, func(w http.ResponseWriter, _ *http.Request) {
+		chunk := make([]byte, writeSize)
+		for range largeBodySize / writeSize {
+			_, err := w.Write(chunk)
+			if err != nil {
+				return
+			}
+		}
+	})
+	addr, out := startProxy(t, app.URL, map[string]string{
+		clientResponseTimeout: shortTimeoutSetting,
+	})
+
+	start := time.Now()
+
+	// The client asks, and never reads the answer.
+	conn := dial(t, addr)
+	send(t, conn, "GET /large HTTP/1.1\r\nHost: app\r\n\r\n")
+
+	line := out.requestLine(t)
+	wantTimedOut(t, start)
+	wantLine(t, line, http.StatusOK, requestlog.ActionTimedOut)
+}

internal/requestlog/requestlog.go

@@ -0,0 +1,102 @@
+// Package requestlog writes the lines smallwebwaf prints on stdout: one
+// JSON object per request, marked "type":"request", and the process's own
+// messages as JSON lines marked "type":"process".
+package requestlog
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"log/slog"
+	"time"
+)
+
+// The action a request line names: what smallwebwaf did with the
+// request.
+const (
+	// ActionForward is a request passed to the app.
+	ActionForward = "forward"
+	// ActionTooLarge is a request or response over its size limit.
+	ActionTooLarge = "too_large"
+	// ActionTimedOut is a request or response that ran out of time.
+	ActionTimedOut = "timed_out"
+	// ActionUpstreamError is a request the app could not be reached
+	// for, or whose answer could not be passed on.
+	ActionUpstreamError = "upstream_error"
+)
+
+// timeLayout is RFC 3339 with milliseconds.
+const timeLayout = "2006-01-02T15:04:05.000Z07:00"
+
+// Line is one request's line in the request log. The field names are
+// those of the "Request log" section of SPEC.md.
+//
+//nolint:tagliatelle // SPEC.md's request log names its fields in snake_case
+type Line struct {
+	Type           string `json:"type"`
+	Time           string `json:"time"`
+	ClientIP       string `json:"client_ip"`
+	PeerIP         string `json:"peer_ip"`
+	Method         string `json:"method"`
+	Host           string `json:"host"`
+	Path           string `json:"path"`
+	Query          string `json:"query"`
+	Protocol       string `json:"protocol"`
+	Status         int    `json:"status"`
+	UpstreamStatus int    `json:"upstream_status,omitempty"`
+	RequestBytes   int64  `json:"request_bytes"`
+	ResponseBytes  int64  `json:"response_bytes"`
+	Referer        string `json:"referer"`
+	UserAgent      string `json:"user_agent"`
+	Action         string `json:"action"`
+	// Aborted is true when the client went away early.
+	Aborted bool `json:"aborted,omitempty"`
+	// DurationTotal and DurationUpstreamTotal are in milliseconds.
+	DurationTotal         float64 `json:"duration_total"`
+	DurationUpstreamTotal float64 `json:"duration_upstream_total,omitempty"`
+}
+
+// Write writes line to w as one JSON line marked "type":"request".
+func Write(w io.Writer, line *Line) error {
+	line.Type = "request"
+
+	encoded, err := json.Marshal(line)
+	if err != nil {
+		return fmt.Errorf("encode the request log line: %w", err)
+	}
+
+	_, err = w.Write(append(encoded, '\n'))
+	if err != nil {
+		return fmt.Errorf("write the request log line: %w", err)
+	}
+
+	return nil
+}
+
+// FormatTime formats t for a line's time field: RFC 3339 in UTC, with
+// milliseconds.
+func FormatTime(t time.Time) string {
+	return t.UTC().Format(timeLayout)
+}
+
+// Milliseconds is d in milliseconds, to the microsecond.
+func Milliseconds(d time.Duration) float64 {
+	return float64(d.Microseconds()) / float64(time.Millisecond/time.Microsecond)
+}
+
+// NewProcessLogger returns the logger for the process's own messages:
+// JSON lines on w, marked "type":"process", with the time in the same form
+// as a request line's.
+func NewProcessLogger(w io.Writer) *slog.Logger {
+	handler := slog.NewJSONHandler(w, &slog.HandlerOptions{
+		ReplaceAttr: func(groups []string, attr slog.Attr) slog.Attr {
+			if attr.Key == slog.TimeKey && len(groups) == 0 {
+				return slog.String(slog.TimeKey, FormatTime(attr.Value.Time()))
+			}
+
+			return attr
+		},
+	})
+
+	return slog.New(handler).With("type", "process")
+}

internal/requestlog/requestlog_test.go

@@ -0,0 +1,88 @@
+package requestlog_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"strings"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+func TestWriteWritesOneJSONLineMarkedRequest(t *testing.T) {
+	t.Parallel()
+
+	var out bytes.Buffer
+
+	err := requestlog.Write(&out, &requestlog.Line{
+		Time:          requestlog.FormatTime(time.Date(2026, 10, 3, 12, 0, 0, 0, time.UTC)),
+		ClientIP:      "203.0.113.9",
+		Status:        200,
+		Action:        requestlog.ActionForward,
+		DurationTotal: requestlog.Milliseconds(1500 * time.Microsecond),
+	})
+	if err != nil {
+		t.Fatalf("write: %v", err)
+	}
+
+	text := out.String()
+	if strings.Count(text, "\n") != 1 || !strings.HasSuffix(text, "\n") {
+		t.Fatalf("wrote %q, want one line", text)
+	}
+
+	var fields map[string]any
+
+	err = json.Unmarshal(out.Bytes(), &fields)
+	if err != nil {
+		t.Fatalf("decode %q: %v", text, err)
+	}
+
+	want := map[string]any{
+		"type": "request", "time": "2026-10-03T12:00:00.000Z",
+		"client_ip": "203.0.113.9", "status": 200.0, "action": "forward",
+		"duration_total": 1.5,
+	}
+	for name, value := range want {
+		if fields[name] != value {
+			t.Errorf("%s is %v, want %v", name, fields[name], value)
+		}
+	}
+
+	unset := []string{"upstream_status", "aborted", "duration_upstream_total"}
+	for _, name := range unset {
+		_, present := fields[name]
+		if present {
+			t.Errorf("%s is there with no value to give", name)
+		}
+	}
+}
+
+func TestProcessLinesAreMarkedProcess(t *testing.T) {
+	t.Parallel()
+
+	var out bytes.Buffer
+
+	requestlog.NewProcessLogger(&out).Info("starting", "version", "v1")
+
+	var fields map[string]any
+
+	err := json.Unmarshal(out.Bytes(), &fields)
+	if err != nil {
+		t.Fatalf("decode %q: %v", out.String(), err)
+	}
+
+	if fields["type"] != "process" || fields["msg"] != "starting" ||
+		fields["level"] != "INFO" || fields["version"] != "v1" {
+		t.Errorf("process line %v", fields)
+	}
+
+	timeText, _ := fields["time"].(string)
+
+	logged, err := time.Parse(time.RFC3339, timeText)
+	if err != nil || !strings.HasSuffix(timeText, "Z") ||
+		len(timeText) != len("2006-01-02T15:04:05.000Z") ||
+		time.Since(logged) > time.Minute {
+		t.Errorf("process line time %q, want now in UTC with milliseconds", timeText)
+	}
+}

internal/smallwebwaf/smallwebwaf.go

@@ -0,0 +1,130 @@
+// Package smallwebwaf runs the smallwebwaf process: it reads the settings,
+// serves requests until it is told to stop, and then stops in an orderly
+// way.
+package smallwebwaf
+
+import (
+	"context"
+	"errors"
+	"io"
+	"log/slog"
+	"net"
+	"net/http"
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/config"
+	"sneak.berlin/go/smallwebwaf/internal/proxy"
+	"sneak.berlin/go/smallwebwaf/internal/requestlog"
+)
+
+// shutdownTimeout is how long requests in progress may take to finish
+// once smallwebwaf is told to stop, before their connections are closed.
+// runit and docker wait a little longer before they kill the process.
+const shutdownTimeout = 5 * time.Second
+
+// Params are what Run needs from the process.
+type Params struct {
+	// Version is the version of the binary, set when it is built.
+	Version string
+	// LookupEnv reads an environment variable, normally os.LookupEnv.
+	LookupEnv func(string) (string, bool)
+	// Stdout receives the request log and the process's own messages.
+	Stdout io.Writer
+}
+
+// Main runs smallwebwaf until SIGTERM or SIGINT, and returns the
+// process's exit status.
+func Main(version string) int {
+	ctx, stop := signal.NotifyContext(context.Background(),
+		syscall.SIGTERM, os.Interrupt)
+	defer stop()
+
+	return Run(ctx, Params{
+		Version:   version,
+		LookupEnv: os.LookupEnv,
+		Stdout:    os.Stdout,
+	})
+}
+
+// Run reads the settings, then serves requests until ctx is done. It
+// returns the process's exit status, 1 when smallwebwaf cannot start.
+func Run(ctx context.Context, params Params) int {
+	processLog := requestlog.NewProcessLogger(params.Stdout)
+
+	cfg, err := config.FromEnvironment(params.LookupEnv)
+	if err != nil {
+		processLog.Error("invalid setting", "error", err.Error())
+
+		return 1
+	}
+
+	listener, err := (&net.ListenConfig{}).Listen(ctx, "tcp", cfg.ListenAddr)
+	if err != nil {
+		processLog.Error("cannot listen on SWWAF_LISTEN_ADDR",
+			"error", err.Error())
+
+		return 1
+	}
+
+	server := proxy.New(proxy.Params{
+		Config:     cfg,
+		RequestLog: params.Stdout,
+		ProcessLog: processLog,
+	})
+
+	processLog.Info("starting",
+		"version", params.Version,
+		"address", listener.Addr().String(),
+		"settings", cfg)
+
+	return serve(ctx, server, listener, processLog)
+}
+
+// serve serves requests on listener until ctx is done, then gives the
+// requests in progress shutdownTimeout to finish.
+func serve(
+	ctx context.Context, server *http.Server, listener net.Listener,
+	processLog *slog.Logger,
+) int {
+	served := make(chan error, 1)
+
+	go func() {
+		served <- server.Serve(listener)
+	}()
+
+	select {
+	case err := <-served:
+		processLog.Error("serving failed", "error", err.Error())
+
+		return 1
+	case <-ctx.Done():
+	}
+
+	processLog.Info("stopping")
+
+	shutdownCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx),
+		shutdownTimeout)
+	defer cancel()
+
+	err := server.Shutdown(shutdownCtx)
+	if err != nil {
+		processLog.Warn("requests still in progress were cut off",
+			"error", err.Error())
+
+		_ = server.Close()
+	}
+
+	err = <-served
+	if !errors.Is(err, http.ErrServerClosed) {
+		processLog.Error("serving failed", "error", err.Error())
+
+		return 1
+	}
+
+	processLog.Info("stopped")
+
+	return 0
+}

internal/smallwebwaf/smallwebwaf_test.go

@@ -0,0 +1,225 @@
+package smallwebwaf_test
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"io"
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"sneak.berlin/go/smallwebwaf/internal/smallwebwaf"
+)
+
+const (
+	// waitLimit bounds how long a test waits for what should happen.
+	waitLimit = 10 * time.Second
+	// pollInterval is how often a test looks for a line.
+	pollInterval = 10 * time.Millisecond
+	// testVersion is the version the tests give smallwebwaf.
+	testVersion = "test"
+	// localhost is where the tests listen.
+	localhost  = "127.0.0.1"
+	listenAddr = "SWWAF_LISTEN_ADDR"
+)
+
+// output collects what smallwebwaf writes on stdout.
+type output struct {
+	mu  sync.Mutex
+	buf bytes.Buffer
+}
+
+// Write adds lines smallwebwaf writes.
+func (o *output) Write(p []byte) (int, error) {
+	o.mu.Lock()
+	defer o.mu.Unlock()
+
+	return o.buf.Write(p)
+}
+
+// line returns the first line whose field key is value, waiting for it.
+func (o *output) line(t *testing.T, key, value string) map[string]any {
+	t.Helper()
+
+	deadline := time.Now().Add(waitLimit)
+	for time.Now().Before(deadline) {
+		o.mu.Lock()
+		text := o.buf.String()
+		o.mu.Unlock()
+
+		for line := range strings.Lines(text) {
+			var fields map[string]any
+
+			err := json.Unmarshal([]byte(line), &fields)
+			if err != nil {
+				t.Fatalf("output line %q is not JSON: %v", line, err)
+			}
+
+			if fields[key] == value {
+				return fields
+			}
+		}
+
+		time.Sleep(pollInterval)
+	}
+
+	t.Fatalf("no line with %s %q in the output:\n%s", key, value, o.buf.String())
+
+	return nil
+}
+
+// run runs smallwebwaf with the settings in env until ctx is done, and
+// returns its exit status.
+func run(ctx context.Context, env map[string]string, out *output) int {
+	return smallwebwaf.Run(ctx, smallwebwaf.Params{
+		Version: testVersion,
+		LookupEnv: func(name string) (string, bool) {
+			value, ok := env[name]
+
+			return value, ok
+		},
+		Stdout: out,
+	})
+}
+
+func TestInvalidSettingStopsTheStart(t *testing.T) {
+	t.Parallel()
+
+	out := &output{}
+
+	status := run(t.Context(), map[string]string{"SWWAF_REQUEST_MAX_BYTES": "lots"}, out)
+	if status != 1 {
+		t.Errorf("exit status %d, want 1", status)
+	}
+
+	line := out.line(t, "msg", "invalid setting")
+	message, _ := line["error"].(string)
+
+	if line["type"] != "process" || line["level"] != "ERROR" ||
+		!strings.HasPrefix(message, "SWWAF_REQUEST_MAX_BYTES: ") {
+		t.Errorf("start refused with %v", line)
+	}
+}
+
+func TestAddressInUseStopsTheStart(t *testing.T) {
+	t.Parallel()
+
+	taken, err := (&net.ListenConfig{}).Listen(t.Context(), "tcp", localhost+":0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+
+	defer func() {
+		_ = taken.Close()
+	}()
+
+	out := &output{}
+
+	status := run(t.Context(), map[string]string{listenAddr: taken.Addr().String()}, out)
+	if status != 1 {
+		t.Errorf("exit status %d, want 1", status)
+	}
+
+	out.line(t, "msg", "cannot listen on SWWAF_LISTEN_ADDR")
+}
+
+func TestServesUntilToldToStop(t *testing.T) {
+	t.Parallel()
+
+	app := httptest.NewServer(http.HandlerFunc(
+		func(w http.ResponseWriter, _ *http.Request) {
+			_, _ = io.WriteString(w, "hello from the app")
+		}))
+	defer app.Close()
+
+	ctx, stop := context.WithCancel(t.Context())
+	out := &output{}
+	exited := make(chan int, 1)
+
+	go func() {
+		exited <- run(ctx, map[string]string{
+			listenAddr:           localhost + ":0",
+			"SWWAF_UPSTREAM_URL": app.URL,
+		}, out)
+	}()
+
+	starting := out.line(t, "msg", "starting")
+	wantStartingLine(t, starting, app.URL)
+
+	addr, _ := starting["address"].(string)
+	wantGreeting(t, "http://"+addr+"/")
+	out.line(t, "type", "request")
+
+	stop()
+
+	select {
+	case status := <-exited:
+		if status != 0 {
+			t.Errorf("exit status %d, want 0", status)
+		}
+	case <-time.After(waitLimit):
+		t.Fatal("still running after being told to stop")
+	}
+
+	out.line(t, "msg", "stopped")
+}
+
+// wantStartingLine checks that the line at start gives the version and
+// every setting's value.
+func wantStartingLine(t *testing.T, line map[string]any, appURL string) {
+	t.Helper()
+
+	settings, _ := line["settings"].(map[string]any)
+	want := map[string]any{
+		listenAddr:                        localhost + ":0",
+		"SWWAF_UPSTREAM_URL":              appURL,
+		"SWWAF_TRUSTED_PROXIES":           "10.0.0.0/8,172.16.0.0/12,192.168.0.0/16",
+		"SWWAF_CLIENT_REQUEST_TIMEOUT":    "60s",
+		"SWWAF_CLIENT_RESPONSE_TIMEOUT":   "30m",
+		"SWWAF_UPSTREAM_REQUEST_TIMEOUT":  "60s",
+		"SWWAF_UPSTREAM_RESPONSE_TIMEOUT": "30m",
+		"SWWAF_REQUEST_MAX_BYTES":         "100M",
+		"SWWAF_RESPONSE_MAX_BYTES":        "5G",
+	}
+
+	for name, value := range want {
+		if settings[name] != value {
+			t.Errorf("starting line gives %s=%v, want %v", name, settings[name], value)
+		}
+	}
+
+	if line["version"] != testVersion || line["type"] != "process" {
+		t.Errorf("starting line %v", line)
+	}
+}
+
+// wantGreeting checks that a request to url gets the app's answer.
+func wantGreeting(t *testing.T, url string) {
+	t.Helper()
+
+	req, err := http.NewRequestWithContext(t.Context(), http.MethodGet, url,
+		http.NoBody)
+	if err != nil {
+		t.Fatalf("new request: %v", err)
+	}
+
+	transport := &http.Transport{}
+	defer transport.CloseIdleConnections()
+
+	res, err := (&http.Client{Transport: transport}).Do(req)
+	if err != nil {
+		t.Fatalf("request: %v", err)
+	}
+
+	body, err := io.ReadAll(res.Body)
+	_ = res.Body.Close()
+
+	if err != nil || string(body) != "hello from the app" {
+		t.Errorf("got %q (%v), want the app's answer", body, err)
+	}
+}

package.json

@@ -0,0 +1,5 @@
+{
+  "devDependencies": {
+    "prettier": "3.8.1"
+  }
+}

script/bootstrap

@@ -0,0 +1,143 @@
+#!/bin/sh
+# script/bootstrap: install all dependencies needed to build and develop
+# this repo. Idempotent: every install is guarded by a check so already
+# installed tools are skipped. Base tooling comes from nix, apt, brew,
+# or apk (detected in that order); assumes nothing is present. Node is
+# used directly if installed; otherwise it is installed at a pinned
+# version via nvm (installing nvm itself first, from a hash-verified
+# release archive, never curl | sh). Go comes from the package manager,
+# for gofmt in script/fmt and script/fmt-check: the tests and the linter
+# run in docker and need no Go on the host.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# Pinned versions, 2026-07-06
+NODE_VERSION="22.17.0"
+NVM_VERSION="0.40.3"
+# sha256 of https://github.com/nvm-sh/nvm/archive/refs/tags/v0.40.3.tar.gz
+NVM_SHA256="5f4d6aaa04a177dc93c985e31dbc411ab6b8c6e1e21d8015dbc1372625fcd1d0"
+YARN_VERSION="1.22.22"
+
+PKGMGR=""
+SUDO=""
+
+detect_pkgmgr() {
+    [ -n "$PKGMGR" ] && return 0
+    if command -v nix-env >/dev/null 2>&1; then
+        PKGMGR="nix"
+    elif command -v apt-get >/dev/null 2>&1; then
+        PKGMGR="apt"
+    elif command -v brew >/dev/null 2>&1; then
+        PKGMGR="brew"
+    elif command -v apk >/dev/null 2>&1; then
+        PKGMGR="apk"
+    else
+        echo "bootstrap: no supported package manager (nix, apt, brew, apk)" >&2
+        exit 1
+    fi
+    if [ "$PKGMGR" = "apt" ]; then
+        export DEBIAN_FRONTEND=noninteractive
+        if [ "$(id -u)" != "0" ]; then
+            SUDO="sudo"
+        fi
+    fi
+}
+
+# pkg_install <nix-attr> <apt-pkg> <brew-formula> <apk-pkg>
+pkg_install() {
+    detect_pkgmgr
+    case "$PKGMGR" in
+        nix) nix-env -iA "nixpkgs.$1" ;;
+        apt)
+            # A fresh CI runner may carry no package lists at all.
+            $SUDO env DEBIAN_FRONTEND=noninteractive apt-get update -q
+            $SUDO env DEBIAN_FRONTEND=noninteractive apt-get install -y "$2"
+            ;;
+        brew) brew install "$3" ;;
+        apk) apk add --no-cache "$4" ;;
+    esac
+}
+
+missing() {
+    ! command -v "$1" >/dev/null 2>&1
+}
+
+# verify_sha256 <file> <expected-hash>
+verify_sha256() {
+    if command -v sha256sum >/dev/null 2>&1; then
+        actual="$(sha256sum "$1" | cut -d' ' -f1)"
+    else
+        actual="$(shasum -a 256 "$1" | cut -d' ' -f1)"
+    fi
+    if [ "$actual" != "$2" ]; then
+        echo "bootstrap: sha256 mismatch for $1" >&2
+        echo "  expected: $2" >&2
+        echo "  actual:   $actual" >&2
+        exit 1
+    fi
+}
+
+# nvm is a bash script; run a command in a bash with nvm loaded
+nvm_sh() {
+    bash -c ". \"\$HOME/.nvm/nvm.sh\" && $*"
+}
+
+ensure_nvm() {
+    [ -s "$HOME/.nvm/nvm.sh" ] && return 0
+    # nvm prerequisites; nvm itself requires bash
+    if missing bash; then pkg_install bash bash bash bash; fi
+    if missing curl; then pkg_install curl curl curl curl; fi
+    if missing git; then pkg_install git git git git; fi
+    tmp="$(mktemp -d)"
+    curl -fsSL -o "$tmp/nvm.tar.gz" \
+        "https://github.com/nvm-sh/nvm/archive/refs/tags/v${NVM_VERSION}.tar.gz"
+    verify_sha256 "$tmp/nvm.tar.gz" "$NVM_SHA256"
+    mkdir -p "$HOME/.nvm"
+    tar -xzf "$tmp/nvm.tar.gz" -C "$HOME/.nvm" --strip-components=1
+    rm -rf "$tmp"
+}
+
+ensure_node() {
+    if ! missing node; then return 0; fi
+    ensure_nvm
+    nvm_sh "nvm install $NODE_VERSION"
+}
+
+ensure_yarn() {
+    if ! missing yarn; then return 0; fi
+    if ! missing corepack; then
+        corepack enable
+        corepack prepare "yarn@$YARN_VERSION" --activate
+    elif [ -s "$HOME/.nvm/nvm.sh" ]; then
+        nvm_sh "nvm use $NODE_VERSION >/dev/null && corepack enable && \
+            corepack prepare yarn@$YARN_VERSION --activate"
+    else
+        npm install -g "yarn@$YARN_VERSION"
+    fi
+}
+
+install_js_deps() {
+    if missing yarn && [ -s "$HOME/.nvm/nvm.sh" ]; then
+        nvm_sh "nvm use $NODE_VERSION >/dev/null && cd \"$ROOT\" && \
+            yarn install --frozen-lockfile"
+    else
+        yarn install --frozen-lockfile
+    fi
+}
+
+main() {
+    cd "$ROOT"
+
+    if missing make; then pkg_install gnumake make make make; fi
+    if missing git; then pkg_install git git git git; fi
+    if missing gofmt; then pkg_install go golang go go; fi
+
+    ensure_node
+    ensure_yarn
+    install_js_deps
+
+    echo "bootstrap complete"
+}
+
+main "$@"

script/build

@@ -0,0 +1,18 @@
+#!/bin/sh
+# script/build: build bin/smallwebwaf on the host, with Go installed, for
+# working on the code by hand. The version it reports comes from git, as
+# in script/docker.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    go build -trimpath -ldflags "-X main.Version=$version" \
+        -o bin/smallwebwaf ./cmd/smallwebwaf
+}
+
+main "$@"

script/check

@@ -0,0 +1,16 @@
+#!/bin/sh
+# script/check: run all checks (test, lint, fmt-check). Our own
+# extension to scripts-to-rule-them-all. test and lint are Docker
+# phases; fmt-check is native, because a formatter writes the working
+# tree. Must not modify any files.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+
+main() {
+    "$SCRIPT_DIR/test"
+    "$SCRIPT_DIR/lint"
+    "$SCRIPT_DIR/fmt-check"
+}
+
+main "$@"

script/cibuild

@@ -0,0 +1,29 @@
+#!/bin/sh
+# script/cibuild: run the CI build. It bootstraps first: a CI runner
+# checks out and runs this and nothing else, and script/fmt-check runs
+# the formatter on the host, which a pristine checkout cannot do.
+# --no-cache for the same reason as script/docker: the gate phases the
+# final stage depends on are RUN steps, and a cached one is a check that
+# did not run.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    "$SCRIPT_DIR/bootstrap"
+    "$SCRIPT_DIR/check"
+    # Own line: a failing command substitution inside an argument does
+    # not trip `set -e`, so the inline form degrades silently to an
+    # empty constant. VERSION is computed here because .dockerignore
+    # excludes .git, so `git describe` in a build stage yields an empty
+    # version without failing.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$("$SCRIPT_DIR/projectname")" .
+}
+
+main "$@"

script/docker

@@ -0,0 +1,25 @@
+#!/bin/sh
+# script/docker: build the Docker image tagged with the project name.
+# Identical in all repos; the tag comes from script/projectname.
+# --no-cache because the gate phases the final stage depends on are RUN
+# steps, and a cached one is a check that did not run.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    # Own line: a failing command substitution inside an argument does
+    # not trip `set -e`, so the inline form degrades silently to an
+    # empty constant. VERSION is computed here because .dockerignore
+    # excludes .git, so `git describe` in a build stage yields an empty
+    # version without failing.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$("$SCRIPT_DIR/projectname")" .
+}
+
+main "$@"

script/fmt

@@ -0,0 +1,33 @@
+#!/bin/sh
+# script/fmt: format all files (writes): the Go code with gofmt, the
+# Markdown with prettier.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# Must match the pin in script/bootstrap.
+NODE_VERSION="22.17.0"
+
+# script/bootstrap installs node and yarn under nvm and leaves neither
+# on the PATH of the shell that called it, so resolve the pinned
+# toolchain here the way bootstrap's own install step does. nvm is a
+# bash script, hence the subshell.
+run_yarn() {
+    if command -v yarn >/dev/null 2>&1; then
+        exec yarn "$@"
+    fi
+    if [ ! -s "$HOME/.nvm/nvm.sh" ]; then
+        echo "fmt: no yarn; run script/bootstrap first" >&2
+        exit 1
+    fi
+    exec bash -c '. "$HOME/.nvm/nvm.sh" && nvm use "$1" >/dev/null &&
+        shift && exec yarn "$@"' bash "$NODE_VERSION" "$@"
+}
+
+main() {
+    cd "$ROOT"
+    gofmt -w cmd internal
+    run_yarn run prettier --write '**/*.md' --tab-width 4 --prose-wrap always
+}
+
+main "$@"

script/fmt-check

@@ -0,0 +1,38 @@
+#!/bin/sh
+# script/fmt-check: check formatting (read-only): the Go code with
+# gofmt, the Markdown with prettier.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+# Must match the pin in script/bootstrap.
+NODE_VERSION="22.17.0"
+
+# script/bootstrap installs node and yarn under nvm and leaves neither
+# on the PATH of the shell that called it, so resolve the pinned
+# toolchain here the way bootstrap's own install step does. nvm is a
+# bash script, hence the subshell.
+run_yarn() {
+    if command -v yarn >/dev/null 2>&1; then
+        exec yarn "$@"
+    fi
+    if [ ! -s "$HOME/.nvm/nvm.sh" ]; then
+        echo "fmt-check: no yarn; run script/bootstrap first" >&2
+        exit 1
+    fi
+    exec bash -c '. "$HOME/.nvm/nvm.sh" && nvm use "$1" >/dev/null &&
+        shift && exec yarn "$@"' bash "$NODE_VERSION" "$@"
+}
+
+main() {
+    cd "$ROOT"
+    unformatted="$(gofmt -l cmd internal)"
+    if [ -n "$unformatted" ]; then
+        echo "fmt-check: gofmt would change:" >&2
+        echo "$unformatted" >&2
+        exit 1
+    fi
+    run_yarn run prettier --check '**/*.md' --tab-width 4 --prose-wrap always
+}
+
+main "$@"

script/install-precommit

@@ -0,0 +1,16 @@
+#!/bin/sh
+# script/install-precommit: install the git pre-commit hook that runs
+# script/precommit. Our own extension to scripts-to-rule-them-all.
+set -eu
+
+ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    hook=".git/hooks/pre-commit"
+    printf '#!/bin/sh\nset -e\nscript/precommit\n' > .git/hooks/pre-commit
+    chmod +x .git/hooks/pre-commit
+    echo "pre-commit hook installed: runs script/precommit"
+}
+
+main "$@"

script/lint

@@ -0,0 +1,23 @@
+#!/bin/sh
+# script/lint: run the linter. Linting is a phase of the Dockerfile and
+# this builds that phase alone; the linter is never installed or run on
+# a developer host, where a shared result cache and a host-global lock
+# make its answer untrustworthy.
+#
+# The phase is not the last stage in the file, so it is built only when
+# --target names it. --no-cache because a cached lint layer is a lint
+# that did not run. The tag makes each build replace the previous image
+# instead of leaving a dangling one behind.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    docker build --no-cache \
+        --target lint \
+        -t "$("$SCRIPT_DIR/projectname")-lint" .
+}
+
+main "$@"

script/precommit

@@ -0,0 +1,12 @@
+#!/bin/sh
+# script/precommit: run by the git pre-commit hook; fails the commit if
+# checks fail. Our own extension to scripts-to-rule-them-all.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+
+main() {
+    "$SCRIPT_DIR/check"
+}
+
+main "$@"

script/projectname

@@ -0,0 +1,12 @@
+#!/bin/sh
+# script/projectname: output the name of this project. Our own
+# extension to scripts-to-rule-them-all. Other scripts that need the
+# name (e.g. script/docker) call this, so they can stay identical
+# across all repos.
+set -eu
+
+main() {
+    echo "smallwebwaf"
+}
+
+main "$@"

script/run

@@ -0,0 +1,14 @@
+#!/bin/sh
+# script/run: build bin/smallwebwaf with script/build and run it, with
+# the settings in the environment.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    "$SCRIPT_DIR/build"
+    exec "$ROOT/bin/smallwebwaf"
+}
+
+main "$@"

script/setup

@@ -0,0 +1,13 @@
+#!/bin/sh
+# script/setup: set up the repo for development after a fresh clone:
+# installs dependencies and the git pre-commit hook.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+
+main() {
+    "$SCRIPT_DIR/bootstrap"
+    "$SCRIPT_DIR/install-precommit"
+}
+
+main "$@"

script/test

@@ -0,0 +1,19 @@
+#!/bin/sh
+# script/test: run the test suite. Testing is a phase of the Dockerfile
+# and this builds that phase alone, on the same terms as script/lint:
+# --target because a phase that is not the last stage is built only when
+# named, --no-cache because a cached test layer is a test that did not
+# run, and a tag so each build replaces the previous image.
+set -eu
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd -P)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd -P)"
+
+main() {
+    cd "$ROOT"
+    docker build --no-cache \
+        --target test \
+        -t "$("$SCRIPT_DIR/projectname")-test" .
+}
+
+main "$@"

yarn.lock

@@ -0,0 +1,8 @@
+# THIS IS AN AUTOGENERATED FILE. DO NOT EDIT THIS FILE DIRECTLY.
+# yarn lockfile v1
+
+
+prettier@3.8.1:
+  version "3.8.1"
+  resolved "https://registry.yarnpkg.com/prettier/-/prettier-3.8.1.tgz#edf48977cf991558f4fcbd8a3ba6015ba2a3a173"
+  integrity sha512-UOnG6LftzbdaHZcKoPFtOcCKztrQ57WkHDeRD9t/PTQtmT0NHSeWWepj6pS0z/N7+08BHFDQVUrfmfMRcZwbMg==