468fd46 · favela-markdown

Changed files (3) +27 -4

spec/design.md +2 -2

…	…	@@ -211,8 +211,8 @@ mismatch class that caused the 2026-07-31 incidents.
211	211	markdown image syntax. `record-vocabulary.md` §11 #2, still open. **The image/video downsampling
212	212	work belongs here**, not in a composer: blob bytes are content-addressed and sealed into signed
213	213	projections, so media policy is an authoring-time decision of the native record.
214		2. Threading natively — `{root, parent}` is settled as canonical (§3.3), but its markdown-native
215		spelling is not.
	214	2. ~~Threading natively~~ CLOSED 2026-09-05 — `record-vocabulary.md` §3.3: frontmatter,
	215	room-relative `<collection>/<position>`, and no CID because the record is mutable.
216	216	3. Community/list URIs for `visibility` values (§11 #4).
217	217	4. Reply visibility inheritance (§11 #5) — almost certainly inherit, but the failure mode is
218	218	publishing to the wrong people, so it must be explicit.

spec/library.md +2 -1

…	…	@@ -323,7 +323,8 @@ OrgEventLog, and the autobee migration guard. It converts a silent fork into a l
323	323	1. Embeds and attachments across atproto blob refs, AS2 `attachment` and markdown image syntax
324	324	(`record-vocabulary.md` §11 #2). The image/video downsampling work belongs here — media policy is
325	325	an authoring-time decision of the native record, not a composer concern.
326		2. Threading's markdown-native spelling. `{root, parent}` is settled as canonical (§3.3); how it
	326	2. ~~Threading's markdown-native spelling~~ CLOSED 2026-09-05 — `record-vocabulary.md` §3.3.
	327	Superseded text: `{root, parent}` is settled as canonical (§3.3); how it
327	328	is written in a file is not.
328	329	3. Community and list URIs for `visibility` (§11 #4).
329	330	4. Reply visibility inheritance (§11 #5) — almost certainly inherit, but the failure mode is

spec/record-vocabulary.md +23 -1

…	…	@@ -179,7 +179,29 @@ The old reasoning was that HTML → byte-offset facets is not reliably recoverab
179	179	relevant: nothing recovers facets from HTML, because facets are produced from the markdown parse,
180	180	which yields exact source ranges (§3.4). Storing them was solving a problem that the parse removes.
181	181
182		Threading: `{root, parent}` is canonical, `inReplyTo` is derived. Unchanged — atproto's pair
	182	Threading: `{root, parent}` is canonical, `inReplyTo` is derived. Its spelling is settled
	183	2026-09-05:
	184
	185	```json
	186	"reply": { "root": "net.favela.post/3msdnle65lk22", "parent": "net.favela.post/3msdz9k2j1a04" }
	187	```
	188
	189	- Frontmatter, not body syntax. A reply target is metadata, and putting it in prose would make
	190	addressing parseable from loose text — the thing §4.1 refuses.
	191	- Room-relative, because a reply is written in the room its parent lives in
	192	(`SOCIAL-MODEL.md` §4.1), so the room segment would be redundant in every native reply. Collection
	193	is kept: a room holds more than one record collection.
	194	- ⛔ No CID. atproto's `strongRef` pins `{uri, cid}`; ours cannot, because §0.1 makes the record
	195	mutable and the hash changes on every edit. A reply must survive its parent being edited, so it
	196	names the parent's identity, never a version of it. The projection supplies the CID from the
	197	parent's stored projection (§7.2) — available precisely because we stopped deriving.
	198	- A foreign parent carries its origin URI as observed (`interaction.md` §3.2) —
	199	`at://…` or `https://…` — since there is no room-relative form for something that lives elsewhere.
	200
	201	⚠ There is no cross-room reply. Replying means writing in the parent's room; if you cannot write
	202	there, you cannot reply. That is not a restriction added here, it is what §4.1 already means.
	203
	204	Unchanged — atproto's pair
183	205	carries strictly more than AS2's single URI, and the projection down is trivial. Its markdown-native
184	206	spelling is still open (`design.md` §8 #2).
185	207

favela-markdown commits / 468fd46

Changed files (3) +27 -4