You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Release-notes dates drift between Month D, YYYY and ISO 8601, because the format is not written down anywhere and nothing in CI checks it.
Background
#419 normalized three dates in docs/_openvox-server_8x/release_notes.markdown from Released 2026-07-22. to Released July 22, 2026..
Before that PR, the repo stood at 19 spelled-out dates against 3 ISO ones, and the openvox-server release notes page was internally inconsistent, carrying both formats on the same page. The ISO dates were not a single bad import. They arrived on at least two separate occasions three months apart: fa735cddf (2026-05-04, the 8.13.0 date) and again with the 8.15.x notes in July.
So this will recur the next time someone adds a release stanza.
Why the spelled-out form
CONTRIBUTING.md says to follow the Google developer documentation style guide, which calls for January 1, 2016 and specifically discourages all-numeric dates as ambiguous across locales. That is also what 19 of the 22 existing dates already use.
The gap is that CONTRIBUTING.md never says this explicitly. It only defers to Google's guide in general, which is easy to miss when you are pattern-matching off a nearby release stanza that happens to use the other format.
Proposed fix
Two options, not mutually exclusive:
1. Document it. Add a line under ### Grammar and spelling in CONTRIBUTING.md:
Dates: use Month D, YYYY (May 4, 2026); avoid all-numeric formats, which are ambiguous across locales
2. Enforce it. A check would catch this at PR time rather than at review time. The pattern is narrow enough to match reliably:
^Released [0-9]{4}-[0-9]{2}-[0-9]{2}
CI already runs both markdownlint-cli2 and rake, so this could live in either:
a markdownlint custom rule (customRules is not currently configured in .markdownlint-cli2.yaml, so this would be the first)
or a small rake task alongside the existing test:links
Option 1 alone is cheap and probably worth doing regardless. Option 2 is what actually prevents recurrence.
Summary
Release-notes dates drift between
Month D, YYYYand ISO 8601, because the format is not written down anywhere and nothing in CI checks it.Background
#419 normalized three dates in
docs/_openvox-server_8x/release_notes.markdownfromReleased 2026-07-22.toReleased July 22, 2026..Before that PR, the repo stood at 19 spelled-out dates against 3 ISO ones, and the openvox-server release notes page was internally inconsistent, carrying both formats on the same page. The ISO dates were not a single bad import. They arrived on at least two separate occasions three months apart:
fa735cddf(2026-05-04, the 8.13.0 date) and again with the 8.15.x notes in July.So this will recur the next time someone adds a release stanza.
Why the spelled-out form
CONTRIBUTING.mdsays to follow the Google developer documentation style guide, which calls forJanuary 1, 2016and specifically discourages all-numeric dates as ambiguous across locales. That is also what 19 of the 22 existing dates already use.The gap is that
CONTRIBUTING.mdnever says this explicitly. It only defers to Google's guide in general, which is easy to miss when you are pattern-matching off a nearby release stanza that happens to use the other format.Proposed fix
Two options, not mutually exclusive:
1. Document it. Add a line under
### Grammar and spellinginCONTRIBUTING.md:2. Enforce it. A check would catch this at PR time rather than at review time. The pattern is narrow enough to match reliably:
CI already runs both
markdownlint-cli2andrake, so this could live in either:customRulesis not currently configured in.markdownlint-cli2.yaml, so this would be the first)raketask alongside the existingtest:linksOption 1 alone is cheap and probably worth doing regardless. Option 2 is what actually prevents recurrence.
Notes
docs/too.