Обсуждение: order of messages on www.p.o/list
Hi, I'm interested to follow PG development but -hackers is too noisy for me to subscribe. I tried following the list on www page few times per day. However, sometimes I refresh the page, and I see that new mails have arrived somewhere in the middle (above the bottom mail), and I don't know which. So I either lose time rereading list of mails (and mails themselves) or otherwise I miss mails which showed up somehere in the middle.. I'm refering to a page like this: https://www.postgresql.org/list/pgsql-hackers/since/201902210000/ Is that list ordered by time accounting for timezone ? For example if someone in -11:00 sends a message at 14:00 UTC, does their message show up above somone else in -04:00 who mails at 13:00 UTC ? Because their "Date:" header says 03:00. Or perhaps it's an issue of the RFC822 "Date:" header vs. the received date? All the headers are just annotations, and so I suggest the page should be ordered not by the "date:" header set by the remote MUA but by the date in the most recent "Received:" header, added by the local MTA. That avoids issue of wrong time setting on remote side. Thanks, Justin
Checking back. This is still an issue. The web list archives seems to sort things by the MUA's "Date" header (??) without normalizing the timezones. When I refresh the page, I expect the newer messages to show up at the bottom. It should be "append-mostly". But with good consistency I see messages shift around in both directions, and it makes this a less useful/efficient way to follow the list. On Thu, Feb 21, 2019 at 12:56:41PM -0600, Justin Pryzby wrote: > Hi, > > I'm interested to follow PG development but -hackers is too noisy for me to > subscribe. I tried following the list on www page few times per day. However, > sometimes I refresh the page, and I see that new mails have arrived somewhere > in the middle (above the bottom mail), and I don't know which. So I either > lose time rereading list of mails (and mails themselves) or otherwise I miss > mails which showed up somehere in the middle.. > > I'm refering to a page like this: > https://www.postgresql.org/list/pgsql-hackers/since/201902210000/ > > Is that list ordered by time accounting for timezone ? For example if someone > in -11:00 sends a message at 14:00 UTC, does their message show up above somone > else in -04:00 who mails at 13:00 UTC ? Because their "Date:" header says > 03:00. > > Or perhaps it's an issue of the RFC822 "Date:" header vs. the received date? > All the headers are just annotations, and so I suggest the page should be > ordered not by the "date:" header set by the remote MUA but by the date in the > most recent "Received:" header, added by the local MTA. That avoids issue of > wrong time setting on remote side.
The date they're sorted by is the result of our forgiving_date_decode() which can be found at https://github.com/postgres/pgarchives/blob/master/loader/lib/parser.py#L424. It definitely at least attempts to normalize timezones, including some of the wonderfully creative ways that MUAs invent new timezones. They will also be capped so that it will never accept dates that are more than 4 hours into the future (which is surprisingly common, because so many MUAs are badly configured) One very common case of messages showing up "not at the end" is simply that they took time to deliver. Either because they were stuck in moderation, or because they were stuck in delivery somewhere along the way. I think that ordering them by the time they arrive at the archive server is simply even more wrong, particularly when considering moderation. //Magnus On Sun, Jan 16, 2022 at 5:49 PM Justin Pryzby <pryzby@telsasoft.com> wrote: > > Checking back. This is still an issue. > > The web list archives seems to sort things by the MUA's "Date" header (??) > without normalizing the timezones. When I refresh the page, I expect the newer > messages to show up at the bottom. It should be "append-mostly". But with > good consistency I see messages shift around in both directions, and it makes > this a less useful/efficient way to follow the list. > > On Thu, Feb 21, 2019 at 12:56:41PM -0600, Justin Pryzby wrote: > > Hi, > > > > I'm interested to follow PG development but -hackers is too noisy for me to > > subscribe. I tried following the list on www page few times per day. However, > > sometimes I refresh the page, and I see that new mails have arrived somewhere > > in the middle (above the bottom mail), and I don't know which. So I either > > lose time rereading list of mails (and mails themselves) or otherwise I miss > > mails which showed up somehere in the middle.. > > > > I'm refering to a page like this: > > https://www.postgresql.org/list/pgsql-hackers/since/201902210000/ > > > > Is that list ordered by time accounting for timezone ? For example if someone > > in -11:00 sends a message at 14:00 UTC, does their message show up above somone > > else in -04:00 who mails at 13:00 UTC ? Because their "Date:" header says > > 03:00. > > > > Or perhaps it's an issue of the RFC822 "Date:" header vs. the received date? > > All the headers are just annotations, and so I suggest the page should be > > ordered not by the "date:" header set by the remote MUA but by the date in the > > most recent "Received:" header, added by the local MTA. That avoids issue of > > wrong time setting on remote side. > >