2025/11/11

Newest at the top

2025-11-11 20:14:16 +0100 <tomsmeding> "Last modified 17/3/00"
2025-11-11 20:14:00 +0100 <EvanR> wow nice css
2025-11-11 20:13:23 +0100 <tomsmeding> EvanR: is this the authoritative reference on what the MJD is? https://core2.gsfc.nasa.gov/time/ Because it defines it as intrinsically linked to a _calendar_ day, with no timezone specified
2025-11-11 20:11:13 +0100 <EvanR> yeah day (unless you're talking NASA MJD stuff) is really a localization thing
2025-11-11 20:10:59 +0100 <tomsmeding> does anyone here remember what the old ircbrowse, when it was still hosted by its original creator (Chris Done), used as the timezone for events?
2025-11-11 20:09:48 +0100 <tomsmeding> for me, being in UTC+1/UTC+2, the UTC day would match closely enough with my calendar day that it's intuitive, but if you're, say, in California or in New Zealand, the ircbrowse "day" is something completely weird
2025-11-11 20:08:55 +0100 <tomsmeding> in a way it's weird in the first place that ircbrowse has the notion of "day", and that day is whatever matches the system timezone on the server it runs on
2025-11-11 20:08:10 +0100 <tomsmeding> without DST I could just leave everything as-is and translate on-the-fly when serving the web page
2025-11-11 20:07:15 +0100 <tomsmeding> DST should be abolished
2025-11-11 20:06:49 +0100haltingsolver(~cmo@2604:3d09:207f:8000:d250:ea0c:366a:6e73)
2025-11-11 20:05:23 +0100 <EvanR> the fascinating problems we put ourselves in
2025-11-11 20:04:44 +0100 <tomsmeding> you know what? I'm going to think about this later and you guys will have to live with the current mess a while longer :p
2025-11-11 20:03:26 +0100 <tomsmeding> it's whether it happened at 01:30 or 02:30 UTC
2025-11-11 20:02:37 +0100tromp(~textual@2001:1c00:3487:1b00:bd50:5f58:be67:a48d)
2025-11-11 20:02:36 +0100 <haskellbridge> <Zemyla> Very few people will care whether a message took place in the summer 1:30 or the winter 1:30.
2025-11-11 20:02:06 +0100 <haskellbridge> <Zemyla> Yeah, you just have to make sure the overlap isn't a problem.
2025-11-11 20:01:45 +0100 <tomsmeding> that's a point
2025-11-11 20:01:40 +0100 <EvanR> flip a coin
2025-11-11 20:01:36 +0100 <EvanR> if there's only 1 message during the ambiguous period, may it's not that important what time it happened xD
2025-11-11 20:01:11 +0100 <tomsmeding> EvanR: possibly
2025-11-11 20:01:04 +0100 <tomsmeding> if there are enough messages you can spot the jump and it's not ambiguous, but if there's, say, only one message in that two-hour slot, it's ambiguous where it is
2025-11-11 20:00:27 +0100 <tomsmeding> so when you switch from summer time to winter time and the clock goes back an hour, you get times between 01:00 and 02:00 twice
2025-11-11 20:00:26 +0100 <EvanR> it would need a multi-part migration which does something to make a translation table of old IDs, reimports, then fixes all the IDs
2025-11-11 19:59:50 +0100 <tomsmeding> another part of the problem is that on a system with system time with DST, znc logs are technically ambiguous since it just contains the HH:MM:SS time for every message
2025-11-11 19:58:57 +0100lucabtz(~lucabtz@user/lucabtz) (Remote host closed the connection)
2025-11-11 19:57:33 +0100 <tomsmeding> Zemyla: so possible, but rather a nuisancve
2025-11-11 19:57:14 +0100 <tomsmeding> I have wanted to rewrite this system for years but I never had the motivation to
2025-11-11 19:56:55 +0100 <tomsmeding> and people might have bookmarks to specific events, which have the ID in the URL
2025-11-11 19:56:41 +0100 <tomsmeding> the issue is this: the import to postgres assigns IDs to the events, and those IDs are not necessarily stable when you do a clean re-import of everything
2025-11-11 19:56:19 +0100 <tomsmeding> ok so the logs are there in the znc log files but the import to postgres borks
2025-11-11 19:55:25 +0100Zemyla(~Zemyla@72.178.108.235) (Quit: Client closed)
2025-11-11 19:52:43 +0100 <tomsmeding> and 2. me not doing that
2025-11-11 19:52:36 +0100 <tomsmeding> and apparently assuming that the server is in UTC
2025-11-11 19:52:31 +0100 <humasect> it will be evaluated when necessary
2025-11-11 19:52:25 +0100 <tomsmeding> well this is a combination of 1. the original author of ircbrowse creating this hideous postgres setup
2025-11-11 19:51:52 +0100 <EvanR> how many haskell programmers does it take to get time and date right
2025-11-11 19:51:48 +0100 <haskellbridge> <Zemyla> I'm pretty sure it's impossible, then.
2025-11-11 19:51:35 +0100tomsmedingchecks if they're in the database
2025-11-11 19:51:18 +0100 <EvanR> nice
2025-11-11 19:51:14 +0100 <tomsmeding> but the website drops them
2025-11-11 19:51:08 +0100 <tomsmeding> well the logs are still there in the plain znc logs
2025-11-11 19:51:00 +0100 <EvanR> we lost logs or we will lose logs
2025-11-11 19:50:49 +0100 <tomsmeding> anyway you can see the opposite jump here https://ircbrowse.tomsmeding.com/day/lchaskell/2025/03/30?id=1517022#trid1517022
2025-11-11 19:50:48 +0100Googulator8(~Googulato@2a01-036d-0106-0180-8127-ba79-55a7-6f29.pool6.digikabel.hu)
2025-11-11 19:50:42 +0100 <sm> no! atlantean calendar
2025-11-11 19:50:38 +0100 <EvanR> you mean if you convert to UTC?
2025-11-11 19:50:37 +0100Googulator8(~Googulato@2a01-036d-0106-0180-8127-ba79-55a7-6f29.pool6.digikabel.hu) (Quit: Client closed)
2025-11-11 19:50:22 +0100 <EvanR> what
2025-11-11 19:50:15 +0100 <tomsmeding> and I now see that means that we lose one hour of logs every october?
2025-11-11 19:49:48 +0100 <EvanR> I vote to convert it to aztec calendar