On Tue, 16 Jun 2009, Tim Connors wrote:
severity 446332 grave
thanks
In trying to track down why imap accounts at work (using UW-imapd, which
uses c-client, which alpine uses), are getting screwed up when being
accessed multiple times, I got reminded of this ancient bug that still
exists.
To reproduce, open up pine on your inbox, open up another pine on your
inbox, and try to mark something as deleted in the first pine session
which is now readonly. It will succeed. If you're lucky, the status line
at top will say "readonly", but it will still succeed to look like it has
deleted a mail. But the changes won't be reflected on disk. In pine, the
requested action (anything other than appending to a mailbox, which uses a
different locking method, at least in the case of MBOX files) should just
be refused. In imap, again the action should be NAKed. It should *never*
be made to look like it succeeded.
Agreed. Are you willing to take the matter upstream? (Is upstream still
alive for UW-IMAP? I know it's barely limping along for alpine.)
And in my case at home using plain old MBOX access, this is not even on
an nfs store - just plain old local filesystems, which should not be
hard to get right (hah! Have you ever traced through the c-client
locking code? Furrfu! At work, it also fails, but of course that's on
nfs, which makes things more tricky especially given the farcical
locking code in c-client).
You might think that not getting your deleted (and new, and read, and
important, etc) flags updated is hardly dataloss, but the CEO is
breathing down my boss's back because his secretary keeps on trying to
delete 800 mails at a time only to have them come back and haunt her
when she restarts her imap mail client in the morning, due to a new
phone system we have which always wants to be connected to the IMAP
store. The user should not be expected to understand that they can only
connect one imap client at a time, especially when that isn't visibly
enforced!
Now, we use a different version of imapd at work, but in tracing the
code, I'm pretty sure the debian package suffers from the same bug
(since the mbox pine bug definitely exists on debian, pine just uses
c-client, and imap also uses c-client, it would make sense for this all
to be just the same bug in c-client), and it'd be nice to at least have
it fixed somewhere (maybe I can win an argument for putting debian on
our mail server at work).
I have to say, you'll win that argument more easily if you don't use the
UW IMAPd. Cyrus or Dovecot are more actively-maintained.
I'll look into these issues in the near term, but that's my advice for
this moment.
-- Asheesh.
--
If rash develops, discontinue use.
--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
