Ack.  This is definitely one of those things that have big consequences.
Maybe it can go into the docs somewhere?

We do have soft quotas, with daily email reminders, up to 7 days.  However,
since faculty and students go on extended vacations here and very often
either don't check their email or ignore the disk quota messages.  It is
also a different behavior from the old way and I'm sure even a gentle
reminder at the beginning of the conversion process won't work.  They won't
remember.  Then I'd have to spend a few hours restoring from backup.

We also don't like to set such a large hard quota because we also have users who never check their email account and I didn't want to have a huge INBOX for users who never use our system.

With the way Unix hard quotas work and the way IMAP works, wouldn't there also be an instance where you can go over your hard quota when you try to move messages from one folder to another? For example, deleting a 2GB INBOX and moving them all to your Trash folder? In those instances, what would happen? Would it fail gracefully and refuse the copy? Or would it mess up the folders?

I was never able to figure out how to suspend delivery for users who are over soft quotas. Do you have any suggestions? I'm using procmail and dmail.

Speaking of restoring, I was able to restore the Seen, Answered fields, but I wasn't able to restore the Forwarded status field. Do you know where I need to look in the .mixstatus file in order to find that flag?

thanks
nancy

Mark Crispin wrote:
Hi Nancy -

I think that you have guessed correctly. Unlike mbx, mix consists of multiple files and doesn't use random access I/O. So, if a hard quota bites when rewriting the status file, the result can be that the update is lost as was the original. Probably, all messages were unread and recent which is basically the same as "no state set".

This happens in mbx and traditional UNIX on some operating systems too. These systems won't let you do a disk write when you are out of quota even if you are just overwriting a disk block that you already have. The result in mbx is a corrupted mbx file.

I don't consider hard disk quotas to be a good means of controlling IMAP usage. The operating systems are too unpredictable and sometimes it just isn't possible to recover from a hard quota exception.

IMHO, a better mechanism is to use soft disk quotas, and then have mechanisms such as suspending new mail delivery to a user who is over soft quota, institute nagging, etc. Then use (much) higher hard quotas to block denial-of-service attacks.

-- 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.


--
-------------------------------------
Nancy Lin                        DECF
1109A Etcheverry Hall    510-642-7291
Office Hours:         2PM-4PM Mon-Thu
-------------------------------------

_______________________________________________
Imap-uw mailing list
Imap-uw@u.washington.edu
https://mailman1.u.washington.edu/mailman/listinfo/imap-uw

Reply via email to