oficina / home Public

Rulings 17-19 as Paul ruled them, and the review's other eight findings

Ruling 18 was wrong and is replaced rather than amended. It said no
activity surface while §0 stands, on the reasoning that the home would
have to learn document names and so retract §0. Fable's review pointed
at `design-activity-pull` — a design on this repo, authored the same
day, where Paul had already ruled a real feed with titles, pulled at
render, holding nothing at rest. Nothing new goes at rest, so §0 is
untouched. The clause that actually changes is §2.2's verb, renders to
stores, and only when that design passes the review it has not had.
§2.2 now carries the pending amendment in place rather than being
silently contradicted, and §13 gains a step 10: blocked on steps 3 and
4, deliberately last, with no claim on anything earlier.

Ruling 17 leaned on a team the data model does not have. §10 had
`teams` as sqid, host and sealed name — one row per instance — so
"Acme" in Docs and "Acme" in Sheets were unrelated rows sharing a DNS
label, and a switcher had no grouping key. §10 now splits `teams`
(suite-wide, carrying the label) from `team_instances` (one per team
and app). Without the split the switcher would have had to group by
the label string, which merges two unrelated teams that picked the
same word.

Ruling 19 follows: a label is reserved suite-wide at enqueue, because
per-app first-come lets a stranger take `acme.sheets` behind your
`acme.docs` on an apex that reads as one suite. §7 states why this is
not the authorization model §2.2 forbids, because that is the
distinction somebody will blur: the home decides who may take a name,
never who may read a document.

Six more from the review. The switcher is named as a zero-JS
`<details>` disclosure of GET links, which rastrillo already ships,
with the selection in the URL rather than a preference row — §14 rules
out preferences beyond theme, and a switcher could have added one
without anyone deciding to. Tiles come from the viewer's own directory
rows, never from every host sharing a label, or a guest in `acme.docs`
would learn Acme uses Sheets, which is gate 4's business. The registry
is compiled-in configuration rather than a table, and only
`integrated: false` entries carry a host, since ruling 16 fixes the
shape of the others.

§3.2's correction overclaimed: what forces exact timestamps is
freshness, not expiry — a day-granular expiry on a long session would
work. And §3.3 now says plainly that coarsening protects a deliberate
publication and nothing else, so the timestamp fingerprint is
unmitigated in a leak, rather than leaving a dead mitigation standing.

§11's count was 129 bare selectors, not 130 — the 130th, `rst-tone`,
only ever appears valued — and the true claim about `Pair`, `Allocate`
and `--rst-header-rule` was dropped along with the false one, so it is
restored.

§11's own ruling is now flagged for re-ruling in §15. Both reasons it
gave for the class vocabulary are dead: the second died at v0.22.0,
the current pin, where the shipped partials flipped to the attribute
spelling and the class spelling in ui/partials/ fell from 124
occurrences to 6. rastrillo §6-v3 rules one spelling and drops the
class selectors before 1.0. This app has no rst- markup yet, so the
switch is free today and a sweep later.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>