91 lines
3.7 KiB
Plaintext
91 lines
3.7 KiB
Plaintext
MBOX-Line: From mrc at CAC.Washington.EDU Wed May 23 23:36:50 2007
|
|
To: imap-protocol@u.washington.edu
|
|
From: Mark Crispin <mrc@CAC.Washington.EDU>
|
|
Date: Fri Jun 8 12:34:39 2018
|
|
Subject: [Imap-protocol] Re: FLAGS response
|
|
In-Reply-To: <1179984406.32181.1730.camel@hurina>
|
|
References: <20070524012401.GA30894@penne.toroid.org>
|
|
<alpine.WNT.0.99.0705231829570.3716@Tomobiki-Cho.CAC.Washignton.EDU>
|
|
<20070524033850.GA31634@penne.toroid.org>
|
|
<alpine.OSX.0.99.0705232059170.10455@pangtzu.panda.com>
|
|
<1179984406.32181.1730.camel@hurina>
|
|
Message-ID: <alpine.OSX.0.99.0705232314230.10455@pangtzu.panda.com>
|
|
|
|
On Thu, 24 May 2007, Timo Sirainen wrote:
|
|
> And I suppose it's OK for server to drop unused keywords from a FLAGS
|
|
> reply and announce that to clients at any point?
|
|
|
|
I would not remove flags, nor change the order of flags in the FLAGS
|
|
response, during a session.
|
|
|
|
> Is it a good idea to drop unused keywords whenever their last user
|
|
> (message) went away?
|
|
|
|
Simple answer: "probably not"
|
|
|
|
See below for more discussion.
|
|
|
|
> Or should I bother adding some kind of a "unused
|
|
> for one month (or so)" timeout too?
|
|
|
|
Simple answer: "probably"
|
|
|
|
More discussion:
|
|
|
|
The IMAP specification is silent on the matter because there was no clear
|
|
understanding on anyone's part what is right.
|
|
|
|
I think that it is alright (from a protocol point of view) to remove
|
|
unused flags in a new session (but NOT during a session!). That's one
|
|
reason why flags are announced; to show what is there.
|
|
|
|
Put another way: a subsequent FLAGS response in a session should only add
|
|
flags at the end of the list, but a FLAGS response in a new session can
|
|
reorder/remove flags.
|
|
|
|
From a usability point of view (as opposed to protocol point of view), I
|
|
think that unused flags should be retained for a reasonable period of
|
|
time, and perhaps not removed at all unless there is some good reason
|
|
(such as an excessive number of unused flags).
|
|
|
|
For example, UW imapd has a limit (inherited from the 1980s when it was
|
|
quite generous) of 30 unique keywords per mailbox. That limit is quite
|
|
inadequate today, especially if people want to do a gmail style mail
|
|
organization. I plan to remove this limit for mix format mailboxes (older
|
|
legacy formats will keep the limit), sooner rather than later.
|
|
|
|
Anyway, if the mailbox uses up its 30 keywords, and there's a bunch of
|
|
unused ones, it's reasonable to clean those up. That hasn't been done by
|
|
any automated mechanism yet. It's on my to-do list; and even if mix
|
|
repeals the limit there will need to be some sort of garbage collection.
|
|
|
|
A server which has no limit, but finds itself with 3000 keywords in the
|
|
mailbox of which only 8 have been used in the past 3 years, is IMHO quite
|
|
adequately provoked into doing some recycling!
|
|
|
|
I use only 6 keywords myself. I would be *very* unhappy with an IMAP
|
|
server that deleted any of my keywords due to disuse, no matter how long
|
|
the period. A mere 6 keywords is not (should not be) costly to keep
|
|
around!
|
|
|
|
I think the key here is "use some common sense". I can't say what numbers
|
|
are "good" numbers since I think we're all groping towards a "sweet spot".
|
|
6 is "obviously small enough to be left alone", and 3000 is "obviously
|
|
huge enough to clean up"; so the threshold is somewhere in the middle.
|
|
|
|
UW imapd has only heard the first rumblings of "30 isn't enough" (but
|
|
there *are* rumblings). So I think that a "sweet spot" for cleaning up
|
|
unused keywords is somewhere in the 2-digit range.
|
|
|
|
The best thing is to try something that appears reasonable and see.
|
|
|
|
Eventually, there will probably be enough collective community knowledge
|
|
that we can provide more precise guidelines to future implementors.
|
|
|
|
-- Mark --
|
|
|
|
http://panda.com/mrc
|
|
Democracy is two wolves and a sheep deciding what to eat for lunch.
|
|
Liberty is a well-armed sheep contesting the vote.
|
|
|