writeonce/docs/stories/language-runtime-database/done/16-web-framework.md

Iteration 16 — the web framework: a .wo library, HTTP/1.1 behind a proxy

Format: fiberloom product/story-iteration-template. Part of Story — one language, one runtime, one database, one binary.

Inserted 2026-08-18. LANDED 2026-08-19 (branch web-framework): just web-app 14/0 — deps chain, auth middleware, CRUD with @unique 409 / FK-restrict 409 / checked-decode 400, :name captures, 404, pipelined keep-alive, SIGTERM, WAL restart persistence. Two notable as-built facts: (1) connection policy is pipelined-keep-alive/close-when-idle — a probe showed a parked keep-alive connection starves accept on a single-threaded server; (2) the chain exposed and fixed two compiler gaps — owned values stored in containers now MOVE (run/container-owned-move), and a dep's internal use paths resolve dep-relatively.

v1 polish LANDED 2026-08-20 (branch framework-v1, developer directive: ship routing/middleware/req-resp as a polished micro-framework, nothing MVC-scale): registration helpers get/post/put/delete_ (the take-Handler shape, probe-proven — deviation 1 of the 16 plan retired), 405 + Allow on wrong-method path hits, HEAD served as GET with the body suppressed (RFC 9110 §9.3.2), a request-line Logging middleware, set_header; just web-app 16/0. It exposed and fixed a third compiler gap: a Text single-segment interpolation of a place crossed let/assignment boundaries uncopied — copy_place_text now sees through Interp (run/interp-borrowed-field).

Auth-in-core LANDED 2026-08-20 (same branch): http/auth.wo — the MECHANISM per the core doctrine: Authorization parsing, pure-.wo base64 (RFC 4648), constant-time ct_eq, req.principal as the blessed "who is this" slot (Middleware.before now takes mut req to write it), BearerAuth + BasicAuth (WWW-Authenticate challenge) middlewares. Policy stays app-side. Probe matrix 26/26 incl. RFC vectors, ASan-clean; web-app dogfoods BearerAuth (its hand-rolled Auth deleted); just web-app 17/0. The framework README now carries the core CHECKLIST: what is ✅ (parsing, routing, middleware, types, auth, JSON), what is a candidate slice (form/multipart, error-mapping hook, config record), and what parks behind 8/11 by design (streaming, backpressure, per-request cancellation).

Form-encoded bodies LANDED 2026-08-20 (same branch): media_type(req) (content-type lowercased, parameters stripped) and form_values(req) (nil unless the content-type says form; '+' and %XX decoded through the existing query decoder). Probe 7/7 release + ASan; web-app's CreateProduct now accepts form OR JSON into one insert path; just web-app 19/0.

Multipart LANDED 2026-08-20 (same branch): http/multipart.wo — RFC 7578 parts (fields + file parts with filename/mime), quoted or bare boundary, strict malformed-is-nil (missing closing marker, missing content-disposition), part_named accessor; whole-body within BODY_MAX, streaming parks behind 8/11. Probe 13/13 + reuse loop, release + ASan; CreateProduct accepts multipart/form/JSON; just web-app 21/0. It surfaced the RETURN flavor of the interp-of-borrowed-place emitter bug (return "${p.content}" — corruption two requests after the handler); fixed in emit_return, pinned in the same fixture.

The framework is written IN writeonce and imported like any dependency (iteration 15 is the prerequisite). TLS terminates at a reverse proxy — browsers get TLS+ALPN+h2 from nginx/caddy while the framework speaks HTTP/1.1 keep-alive behind it, so no TLS exists anywhere in the toolchain. h2c is the parked successor (after iterations 8/23/11, when multiplexing has a scheduler to pay off on).

Spec exists: 2026-08-18-web-framework-design.md sections B (normative) and C (the parked h2c successor). Plan: 2026-08-19-web-framework.md (6 tasks: types/builders; HTTP/1.1 parse+serve; router+interfaces+App; the web-app storefront; the just web-app gate; docs closeout). Two enabling risks retired before planning: interface-field dispatch (probe) and the route-table owned-move double-free (fixed, pinned by run/container-owned-move).

Goals

• The framework, incubated at docs/examples/writeonce-framework/: an HTTP/1.1 keep-alive server core (extracted from the soak-proven log-watcher mcp pattern), a method+path router with :param captures, Req/Resp records with builder helpers, and the no-function-values handler model — Handler / Middleware structural interfaces dispatched by ICALL, with WO-E205 making a non-conforming handler a compile error and ?Resp middleware short-circuiting on the shipped ?T narrowing.
• The consuming app, docs/examples/web-app/: a small storefront (Product/Order as @table classes, list/show/create routes, one auth middleware, JSON responses) that imports the framework through [deps] — the sample exercises the whole chain: fetch → lock → build → serve → durable data across restart.
• Honest limits stated where users read them: single-threaded blocking (concurrency arrives underneath via iterations 8/11), no chunked encoding, no WebSockets, JSON-first (no templates — the removed UI track stays removed).
• Relationship to iteration 25 recorded in both: service blocks later lower onto this library — compiler sugar over the same router, never a rival stack.

Acceptance Criteria

• Given the web-app sample, when just web-app runs, then deps fetch, the app builds, serves list/show/create with @table persistence across a process restart, answers 404/400 correctly, a deliberately trapping handler returns 500 while the server survives, and SIGTERM stops it cleanly — fd- and resident-flat under the opt-in soak.
• Given a handler class missing handle, when the app compiles, then WO-E205 names the class and interface.
• Given nginx in front (documented config), when a browser hits it, then the browser negotiates h2 with the proxy while the backend link is HTTP/1.1 — proving the TLS/h2 story without TLS in writeonce.
• Given the framework extracted to its own repository, when the app's [deps] URL is updated, then nothing else changes (success criterion 4 of the spec).

Out Of Scope

TLS in the toolchain (proxy-terminated by decision); HTTP/2 + the bytes/buffer type (parked to the h2c successor, after 8/23/11); chunked transfer encoding; WebSockets/SSE; templates/SSR; multipart uploads; performance work beyond the soak's flatness gate (benchmarks belong to 22's measurement backbone).

Proposed Solution

Framework modules http/, router/, app.wo as spec §B lays out; the web-app acceptance script is the gate (curl matrix + restart + stop + soak); just web-app wires fetch-build-serve-verify into one command. Plan follows after this iteration is approved on the board.