| Comment: | Sub-manifest: record why the final-group merge cap is unpinned
The merge fires when the two groups together hold at most MAX, and no test separates that from a strict comparison. Reaching the distinguishing case needs a natural boundary at exactly MAX minus the tail, which hash-driven boundaries do not produce in generated name sets; a search over two parameter values and nine hundred starts found no instance. Pinning it would mean choosing filenames by their leading zero bits rather than generating them. The note records the reasoning at the site so the gap is visible to the next reader. |
|---|---|
| Timelines: | family | ancestors | descendants | both | sub-manifest |
| Files: | files | file ages | folders |
| SHA3-256: |
9897845809fe015a42d33ec28e00d02b |
| User & Date: | dmestas on 2026-07-28 01:49:53.364 |
| Other Links: | branch diff | manifest | tags |
|
2026-07-28
| ||
| 02:47 |
Sub-manifest draft: carry the step-5 cursor correction
Regenerated from the RFC suite after the erratum submanifests-00-s5.1-step5-cursor-advance was applied. Step 5 of the tree walk now advances the referencing frame's cursor before pushing the interior frame; as published it omitted that, so returning from an interior descent re-read the same S card and failed the strictly-greater-than-LAST check, leaving every tree with an interior node unwalkable. The walk already implements the corrected reading. Also picks up an MD5 citation and reflowing in unrelated paragraphs. check-in: f2973c41c3 user: dmestas tags: sub-manifest | |
| 01:49 |
Sub-manifest: record why the final-group merge cap is unpinned
The merge fires when the two groups together hold at most MAX, and no test separates that from a strict comparison. Reaching the distinguishing case needs a natural boundary at exactly MAX minus the tail, which hash-driven boundaries do not produce in generated name sets; a search over two parameter values and nine hundred starts found no instance. Pinning it would mean choosing filenames by their leading zero bits rather than generating them. The note records the reasoning at the site so the gap is visible to the next reader. check-in: 9897845809 user: dmestas tags: sub-manifest | |
| 01:19 |
Sub-manifest: accept in the S-card writer what the parser accepts
The S hash is validated case-insensitively and kept raw, so a parsed card carries whatever spelling it was given. The writer asserted the narrower lower-case spelling, so re-emitting a card the parser had just accepted tripped the assertion. It now asserts the parser's own predicate. Reachable only once something emits, which nothing does yet; found by round-tripping an upper-case hash through create. check-in: 6e64c8976c user: dmestas tags: sub-manifest | |