b3e6708 · favela-markdown

Changed files (2) +33 -1

spec/library.md +21 -1

…	…	@@ -309,7 +309,27 @@ OrgEventLog, and the autobee migration guard. It converts a silent fork into a l
309	309	3. Community and list URIs for `visibility` (§11 #4).
310	310	4. Reply visibility inheritance (§11 #5) — almost certainly inherit, but the failure mode is
311	311	publishing to the wrong people, so it must be explicit rather than emergent.
312		5. ~~Where `STRIP_VERSION` is stamped~~ RESOLVED 2026-09-05: the ROOM's StateBase. The signer
	312	5. ⚠ Where `STRIP_VERSION` is stamped — RESOLVED 2026-09-05, but it is only HALF the answer.
	313	hyper-social #43 (2026-08-11) decided per-record `dialect` versioning and it was never
	314	written. The two are not alternatives:
	315
	316	\| \| guards against \| scope \|
	317	\|---\|---\|---\|
	318	\| `STRIP_VERSION` \| two peers on different strips signing one room ⇒ fork \| concurrency, per room \|
	319	\| per-record `dialect` \| a record authored under v1 being re-stripped under v2 \| history, per record \|
	320
	321	⛔ The projection is derived on read, so every sign re-strips every record. Bump the dialect
	322	and every historical post's `app.bsky.feed.post` silently changes — the whole atproto history
	323	rewrites itself, identically on every peer, which is precisely what `STRIP_VERSION` guarantees and
	324	precisely what makes it insufficient alone. That is `record-vocabulary.md` §8's *"never re-encode a
	325	record you did not author"* — applied to yourself, over time.
	326
	327	Both are needed. A record states the dialect it was authored under; `strip()` takes it as an
	328	argument rather than being a single global function; the room's `STRIP_VERSION` continues to gate
	329	who may sign. ⚠ `lib/strip.js` currently exports `strip(source)` with no dialect parameter, so
	330	this is a signature change, not an addition.
	331
	332	Original resolution follows: the ROOM's StateBase. The signer
313	333	role is per-room — a signer materialises that room's record set — so the version that gates it
314	334	must be per-room too. An under-versioned peer refuses the signer role for that room and can
315	335	still sign others, which is what makes a fleet rollout possible at all: a single global stamp would

spec/markdown-strip.md +12 -0

…	…	@@ -212,6 +212,18 @@ facets that land on the wrong characters only for non-ASCII authors.
212	212
213	213	## 9. Open
214	214
	215	0. ⛔ Per-record `dialect` versioning — decided 2026-08-11 (hyper-social #43), never written.
	216	Every rule in this document is normative for one dialect, and §2.1 says adding an extension is a
	217	spec change "because it changes the bytes of every record that uses it". It does worse than that:
	218	because the projection is derived on read, a dialect bump re-strips every record ever written.
	219	A record must therefore state the dialect it was authored under, and `strip()` must take it as an
	220	argument. See `library.md` §11 #5. This is a signature change to the shipped `strip(source)`.
	221
	222	0b. **"Unknown syntax degrades to literal text" — decided 2026-08-11 (#43), stated but not
	223	normative.** §2 makes the function total and the implementation's fallback branches do emit
	224	literal text, but there is no rule saying so and no vector covering it, so it is untested
	225	behaviour rather than a guarantee.
	226
215	227	1. Grapheme counting for the bsky projection is specified in `record-vocabulary.md` §4.3 but the
216	228	segmentation algorithm is not pinned. It must be, for the same determinism reason.
217	229	2. Embed extraction order (§3.5) is asserted to be source order but not otherwise specified;

favela-markdown commits / b3e6708

Changed files (2) +33 -1