Belt B: roadmap serve, component model, and sandbox (P0–P4)

[briansrls]

Summary

This PR lands belt B (gunbc serve) end-to-end, replacing belt A's interim Node server emit with an in-process HTTP service. It introduces the component concept (dispatch button as the second consumer), the design-register scale axis, and the /sandbox surface as the visual witness. The belt-A roadmap dashboard emit is deleted; gunbhub remains on belt A (out of scope).

Key Changes

Belt B service (gunbc serve)

• roadmap_serve.dag: New in-process HTTP handler for GET / (daily), /roadmap, /sandbox; POST /dispatch (in-process, no shell-out); POST /sandbox/echo (echo any wire label)
• node_http_server_emit.dag: Removed belt-A smuggled-exec dispatch route (BeltDispatchRouteConfig, execFile-gunbc handler, env-var channel); roadmap static route table and all roadmap_dashboard_* emitters deleted; module survives only for gunbhub static-server emit (sole live consumer)

Component concept (P0)

• design/component.dag (new): Generic component model — element structure × role/material bindings × interaction states (idle | requested | terminal-per-wire-arm) × wire-contract binding; component_wire_total cross-checks terminal set against wire vocabulary both directions
• roadmap_component.dag (new): Dispatch button instance; terminal states DERIVED from belt_dispatch_all_status_labels (grounded label set, never hand enum); role binding follows ok/refusal split (figure for ok, boundary for refusal); idle reads text, requested reads text-dim; reserved width is longest label + gutter (grows by derivation if a longer arm lands)

Client behavior (P1)

• roadmap_component.dag: dispatch_client_js_source() emits ONE TsProgram (native .dag, never hand-concatenated JS string); served as external asset /assets/roadmap-dispatch.js (external emission avoids EscapeHtml hazard on arrows); ok-predicate is ts_binop || chain DERIVED from belt_dispatch_ok_status_labels
• roadmap_serve.dag: DispatchAssetHandler serves dispatch_client_js_source() as TextJavascriptUtf8

Presentation (P3)

• roadmap_style.dag: Dispatch button styling integrated; dispatch_reserved_width, dispatch_class_* (requested/ok/refused) imported and used; single authority for state's CSS modifier class (consumed by client script, stylesheet, P4 gallery)
• dispatch_presentation_witness_test.dag (new): Reserved width is label projection; each chip's material is the state's role; located fields (step/detail) surface into title attribute

Sandbox surface (P4)

• roadmap_sandbox.dag (new): /sandbox design surface — register at every depth plus every component in every state, all DERIVED from same authorities live surfaces use; role swatches under both themes (literal material paints inline); component gallery renders dispatch_button_component's states one-for-one; live section is one pressable instance per wire arm that POSTs /sandbox/echo?status=LABEL
• roadmap_serve.dag: SandboxPageHandler (GET /sandbox), SandboxAssetHandler (GET /assets/sandbox.js), SandboxEchoHandler (POST /sandbox/echo?status=LABEL)
• roadmap_sandbox_witness_test.dag (new): Gallery totality is wire vocabulary re-witnessed as coverage; echo answers ANY wire arm with same belt JSON shape production sends; live instances driven by shared client program

Design register scale axis (P2)

• design/scale.dag (new): Spacing, type, font stack, dimension tokens (radius, border width, measure, fixed element extents); value-named (space-8 == 8px, text-14 == 14px) for mechanical P2 migration and review-checkability; scale_var, scale_decl, scale_pair, scale_triple, scale_border helpers; untokened-length census (gunbc.design.theme_transition) excludes :root block
• roadmap_style.dag:

https://claude.ai/code/session_019SfKm4AxRS3e6nX2kA8med

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[briansrls]

Anemia dissolution follow-up (D1–D5)

Rebased onto current main (post-#7096). The P0–P5 modeling is unchanged (accepted as modeled); what follows dissolves the anemic seams the new consumers exposed. Discipline: pure re-grounding — emitted bytes and wire strings are unchanged, each item with its receipt.

D1 — MediaType (extdeps/http/server.dag). The fused HttpContentType enum (media-type × charset flattened into variant names) is deleted. Modeled as RFC 9110 §8.3 says it: MediaType { top_level: <closed RFC 6838 registry enum>, subtype: String, parameters: List<MediaTypeParameter> }, charset a cited IANA parameter row; http_content_type_wire_label → media_type_wire_label serializes over the structure; ApplicationOctetStream = empty parameter list (the asymmetry falls out). Four consumers + five witnesses migrated.

• Receipt: media_type_witnesses pins all five wire strings byte-exact + perturbed-parameter/perturbed-subtype REDs. Forward Content-Type parse is named as the later direction, not built.

D2 — typed HTTP status (same file). belt_dispatch_status and serve_text_response re-grounded on HttpStatus; the bare Int survives only at the ServeWireResponse wire seam.

• Receipt: per-variant codes 200/200/404/409/503/502 pinned unchanged (witness_dispatch_status_total).

D3 — CssLength (new extdeps/languages/css/values.dag, cited CSS Values & Units + Fonts 4). ScaleToken.value: String → typed ScaleValue (CssLength{n,unit} | CssFontStack{families}). The value-encoded token name is derived from the magnitude (length_token(px:4) → "space-4"), so the name↔value fork is unwritable by construction. dispatch_reserved_width's join-with-"ch" migrates to css_ch.

• Receipt: roadmap_css() byte-identical (sha unchanged); lift-parity digest holds.

D4 — exemplar roster (roadmap_belt_actuate + gunbc.roster_registry). belt_dispatch_result_exemplars (a hand variant enumeration) enrolled as a DeclaredFrontier row (reason + dissolve_on).

• Receipt: a reconciliation witness pins the exemplar-derived label set against the independent wire-contract pin, with a missing-variant RED; roster_registry_standard_holds_live stays green.

D5 — typed CSS grain (new extdeps/languages/css/{properties,selectors}.dag). The last stringly medium: CssPropertyName over the register's emitted property set (+ CustomProperty for author vars, a counted RawProperty escape) and SelectorSubject = ClassSelector | ElementSelector | PseudoSuffix | RawSelector. material.CssDecl.prop and BuildRule subject re-grounded; every consumer migrated.

• Receipt: stylesheet byte-identical (lift-parity + roadmap/site register witnesses green); css_grain_witnesses pins the serializers with RED controls. Counted residue: RawSelector = 18 (descendant/compound/functional-pseudo, awaiting the combinator grammar), RawProperty = 1 (accent-study code-spacing). CSS value grammar stays a named arc, out of scope.

Local serve receipt (gunbc serve, scratch port):

request	status	content-type	body
GET /	200	text/html; charset=utf-8	page
GET /sandbox	200	text/html; charset=utf-8	sandbox
POST /sandbox/echo?status=spawn_failed	200	application/json; charset=utf-8	{"node_id":"x","status":"spawn_failed","step":"s","detail":"d"}
POST /dispatch/bogus	404	application/json; charset=utf-8	{"node_id":"bogus","status":"node_not_found"}

Green by execution: page, serve, sandbox, register (roadmap + site + lift-parity), component, presentation, belt (incl. D4), static-site, gunbhub, http-serve-route, media-type, css-grain, roster-registry, wcag/palette/accent, live-deploy (apply/emit/intent/service-ready). Whole-tree dag compile: zero hard errors, zero silent-picks from the grain. (live_deploy_readiness execs ssh, absent in this sandbox — environmental, unrelated to this diff.)

---

Generated by Claude Code

[briansrls]

Review against the P0–P5 brief + the D1–D5 dissolution addendum (operator-requested; this is the verdict that moves the row, per ts-dispatch-lifecycle).

1 · Delivery vs brief — all five D-items delivered, verified by execution

I ran the new suites independently at head 5bcea73e1e: media_type_witnesses PASS, css_grain_witnesses PASS, roadmap_belt_actuate_witnesses (wet, includes the exemplar-roster wall) PASS — on top of the P-phase suites verified in the earlier review round.

• D1 is the exemplar for the citation bar: four distinct authorities (RFC 9110 §8.3.1 grammar, RFC 6838 §4.2 closed top-level, IANA media-types registry, IANA character-sets), the fused HttpContentType enum deleted with media_type_wire_label the sole serializer, octet-stream falling out as an empty parameter list instead of a variant quirk, and the forward read named-not-built. Exactly the decompress-the-string move the brief asked for.
• D2 clean: ServeHttpResponse.status: HttpStatus; the one remaining Int is ServeWireResponse — the declared one-way wire projection with its note saying precisely that. Correct seam.
• D3 done by construction, not witness: scale_token(prefix, px) derives the token name from the value, so the name/value fork is unwritable; CssLength grounds in extdeps.languages.css.values.
• D4 enrolled: belt_dispatch_result_exemplars as DeclaredFrontier with the honest dissolve_on (substrate inhabitance enumeration — fold over a coproduct's arms without a hand list). That's the third time this week the group lesson recursed to a smaller grain; the registry now counts it.
• D5 bounded as briefed: properties/selectors/values modules, RawSelector a counted frontier, the value grammar left as the named arc.

2 · Modeling audit

Nothing anemic introduced; the two structural risks I was watching for did not materialize (no second wire vocabulary, no re-minted length concept — CssLength is the one carrier for both the tokens and the reserved-width projection). Residues, all honestly marked on carriers: the RawSelector census count is real burn-down work; the Content-Type forward read is the named later direction; the byte-identity oracles for D3/D5 are claimed on the ts-ui-model row and corroborated by the register suites staying green.

3 · Planning / disposition

Accepted-with-residue: merge on a green floor. Residue is already carried on carriers (RawSelector count, forward read, inhabitance enumeration) — no new sheet rows owed. The workflow-shape lane (#7118) references this branch as the owner of the presentation half; on merge that seam note stays accurate as written.

🤖 Generated with Claude Code

---

Generated by Claude Code