Re: [Evolution] email notification wish list

[Michael Poole <poole troilus org>]

Jamie Zawinski <jwz jwz org> writes:

[snip]
> However, when talking to POP3 (or, I think, IMAP) you're going to be in
> a world of hurt if there are two different processes talking to the same
> server, so you want all server communication to go through one
> connection.  

For POP3, you're right.  The protocol isn't designed to support
multiple clients accessing the same mailbox concurrently, and so it
isn't likely to work well.

For IMAP, they kind of sort of kept multiple readers in mind.  So
there's support, but it's subject to most of the same compromises as
Unix files -- if you have writer-writer or reader-writer conflicts,
somebody is either denied or surprised about what happens.

There's an RFC (2180) about how to mitigate those problems in IMAP4;
it would be useful to at least consult it to see what the Best Current
Practice for IMAP, even if the end decision is that Evolution's model
isn't compatible with the model that RFC2180 assumes.

As for running multiple copies of a notification program, or sharing
one connection to a mail server between mail notification and mail
reader: If there's an easy way to do it, good.  If there's not, I
don't think it's worth bending over backwards to accomodate.  There
are servers that let you do multiple connections reliably, and there
are other ways to multicast notification information (various instant
messaging protocols; in particular, the Cyrus IMAP server's zephyr
notification service).  Feature creep is useful, but if you're writing
a mail client, tackling architecture and protocol design problems in
other domains is not likely to be useful.

Michael