Identifying mails between an IMAP server and a client application

1
votes

The IMAP spec gives great hopes that the UID value may be used for this:

The unique identifier of a message MUST NOT change during the session, and SHOULD NOT change between sessions. Any change of unique identifiers between sessions MUST be detectable using the UIDVALIDITY mechanism discussed below.

But alas, at least for Microsoft Exchange server, my experience shows that the UID of a message does occasionally/often change between sessions, even though it should not, and this is not detectable using UIDVALIDITY (which always stays the same).

As an example of what actually may happen. I say to the server, got any mail? And the server says, sure, here they are, with ids 4, 5, 7, 8 and 11. (Incrementing but not necessesarily contiguous, so far so well.)

So, next time, I say, got any mail newer than 11? And the server says sure, and gives me 14, 15, 16 and 18, but what has actually happened is that it has reassigned the id of the message previously known as e.g. 5 to e.g. 16! So what looks like a new mail, is actually an old mail with a new id!

This means that 1) "should not" bears no value, 2) Exchange is actually following the spec if we interpret "should not" as "it's up to you" and 3) it may be violating the spec on the UIDVALIDITY part (not quite sure as the spec is a bit vague).

It also means that the UID cannot be trusted as a way to identify messages on the server. If I download message 5, store some information on it in the database, and the server renames it to 16 without my knowing, how can I later find it on the server? And how can I know that 16 is a duplicate of what I have stored as 5?

What is a reliable way of identifying mails on the server knowing that UID cannot be used?

1
emailimap
Are you sure you're looking at UIDs and not Sequence numbers? Sequence numbers are subject to change, and are usually what you get with mod libraries unless you specifically ask for UIDs. Also, are you sure the server has not renumbered the message due to the user editing it (which necessitates replacing it with a new message) - Max
Absolutely sure that I use the UID. I also cannot see how a user would edit a received mail (if Exchange even allows that). - Knut Arne Vedaa

1 Answers

2
votes

Looking at the message-id header field is probably your best bet.

If Exchange does what you say it does, then it violates the spec: It's permitted to reassign UIDs, but if it does it absolutely must change UIDVALIDITY. Which particular version/plugin is this?

(EDIT: I hear that Exchange always increases the UIDs when it does this, which is then not a spec violation. Assigning a new lower UID would be a total no-no, but so long as Exchange always increases the UIDs it's not actually violating the spec. I do find that behaviour a bit smelly though.)