data/releases.json is the file the landing table and every plugin page compute
the shipping state from. It is produced by go run . releases, which reads the
release list of each repository the roster names, and it is the one verb here
that writes into the tree rather than reporting on a file with an authority
elsewhere. Nothing runs it on any cadence, and nothing says when what it holds
has stopped being true.
The record is six days old and the landing page states that date to a reader as
the moment the data was taken:
jq -r '.taken' data/releases.json
2026-08-26
go run . build > /dev/null
grep -o -E '[^<>]*read on [^<>]*' dist/index.html
12 rows, read from the roster this repository carries. What each plugin has published was read on 2026-08-26, and the state below is computed from that rather than declared: a row says what is true before anything is published, and a finished release raises it.
Both run 2026-09-01 at acf49ee. No workflow names the verb:
grep -rln 'releases' .github/workflows/; echo "exit=$?"
exit=1
and the three neighbouring verbs that read something published elsewhere each
have one:
for v in pins tokens roster releases; do
printf '%-9s %s\n' "$v" "$(ls .github/workflows/$v.yml 2>/dev/null || echo 'no workflow')"
done
pins .github/workflows/pins.yml
tokens .github/workflows/tokens.yml
roster .github/workflows/roster.yml
releases no workflow
Both run 2026-09-01 at acf49ee.
The record is out of date today
Taken by running the verb and putting the file back:
go run . releases | tail -1
releases: 12 repository(s) recorded as taken on 2026-09-01, 2 entry(s) moved, written to data/releases.json
git diff data/releases.json | grep -E '^[-+] +"(finished|prereleases)"'
- "finished": 1,
+ "finished": 2,
- "finished": 29,
- "prereleases": 53,
+ "finished": 37,
+ "prereleases": 61,
git checkout -- data/releases.json
Run 2026-09-01 at acf49ee. The two that moved are requests and sso, and
neither move changes the state word the page prints today. That is the argument
for this rather than against it: the move that does change one is the move
nothing here would report either. A repository publishing its first finished
release turns a shell row into a shipping one, and until somebody remembers to
run the verb the page goes on printing the old word beside a date a reader takes
for current.
I have not landed the retake. It is a change to a file no open issue lands, and
#23 gives the second reason: it moves the counts the threshold argued there would
be written against.
What this is not
It is not the comparison #25 waits on. That one compares a copy this repository
pins against a file another repository publishes, and where such a comparison
runs is the open question there. This record has no authority anywhere else - the
verb produces the answer, and main.go says exactly that at the call:
grep -n 'the one verb here that writes into' -A 1 main.go
126: // this tree changes. It is also the one verb here that writes into
127- // the tree, because what it produces is the answer itself rather
Run 2026-09-01 at acf49ee. It is also not a leg of the gate, for the reason the
other three are not, which is written at the same call. So what is absent is the
route that asks for it and reports what it found, and that shape has landed three
times on this board already.
Done when
A route asks for the record on a cadence written down with its reason beside the
two schedules already in the tree; it reds while re-taking the record would
change it; its report names each repository whose counts moved, with what they
moved from and to; it writes into no repository; a run against a record that
reproduces what is published passes and says what it compared rather than saying
nothing; and the report is shown to bite by a record edited to disagree with one
published release list, which reds it, and by the same record restored, which
does not.
data/releases.jsonis the file the landing table and every plugin page computethe shipping state from. It is produced by
go run . releases, which reads therelease list of each repository the roster names, and it is the one verb here
that writes into the tree rather than reporting on a file with an authority
elsewhere. Nothing runs it on any cadence, and nothing says when what it holds
has stopped being true.
The record is six days old and the landing page states that date to a reader as
the moment the data was taken:
Both run 2026-09-01 at
acf49ee. No workflow names the verb:and the three neighbouring verbs that read something published elsewhere each
have one:
Both run 2026-09-01 at
acf49ee.The record is out of date today
Taken by running the verb and putting the file back:
Run 2026-09-01 at
acf49ee. The two that moved arerequestsandsso, andneither move changes the state word the page prints today. That is the argument
for this rather than against it: the move that does change one is the move
nothing here would report either. A repository publishing its first finished
release turns a
shellrow into a shipping one, and until somebody remembers torun the verb the page goes on printing the old word beside a date a reader takes
for current.
I have not landed the retake. It is a change to a file no open issue lands, and
#23 gives the second reason: it moves the counts the threshold argued there would
be written against.
What this is not
It is not the comparison #25 waits on. That one compares a copy this repository
pins against a file another repository publishes, and where such a comparison
runs is the open question there. This record has no authority anywhere else - the
verb produces the answer, and
main.gosays exactly that at the call:Run 2026-09-01 at
acf49ee. It is also not a leg of the gate, for the reason theother three are not, which is written at the same call. So what is absent is the
route that asks for it and reports what it found, and that shape has landed three
times on this board already.
Done when
A route asks for the record on a cadence written down with its reason beside the
two schedules already in the tree; it reds while re-taking the record would
change it; its report names each repository whose counts moved, with what they
moved from and to; it writes into no repository; a run against a record that
reproduces what is published passes and says what it compared rather than saying
nothing; and the report is shown to bite by a record edited to disagree with one
published release list, which reds it, and by the same record restored, which
does not.