On Thu, 2006-05-11 at 08:25 -0700, Aaron Stone wrote: > This drives me bonkers. I would have liked to see new IMAP rfc's go in > the other direction by specifying MORE magic mailboxes.
I think allowing something like KEYWORDs or ANNOTATIONs on folders/mailboxes is the way to go. It'd certainly go a long way to allowing clients to interoperate their magic folders- simply look for the folder with the +SENTBOX keyword on it. Even if vendors picked different things- it wouldn't matter, the user/administrator could fix it up for them. In the meantime, making it possible to make sent/sent items/sentmail/sentbox/etc all the same physical mailbox would be nice. Of course, vendors would like this- because they could give folders "types" and store "kinds" of files in there. Like IPM.Appointments or what have you. > The Courier OUTBOX extension is intended to do something like this. It's > because the Courier lead isn't a protocol nerd, he's an implementer. He wants > it to Just Work (TM), and I definitely feel him on this one. I don't entirely know what the OUTBOX extension is for- I don't know of any clients that support it, and using it with clients that don't is fairly dangerous. Without folder typing, I can't see how any of this is a good idea. Better still: With folder typing being additive- clients could refuse to operate on a mailbox with +TYPEX on it- if they don't understand +TYPEX. The user would have to override the types if they knew what they were doing. Better: vendor/* annotations are dumb. Let's make the annotation/keywords have to be URLs to the specification that describes their behavior. That way, if a client implementer runs across a new keyword, they have the URL immediately- thus making google searches for DELETED actually helpful :) > Theoretically, ACAP might be able to hold some of these "universal > client preferences" but in reality it's the most horrifically > complicated registry system ever. ACAP is a pipe dream. Right now, my IMAP clients store any configuration they need in /netMail-Configuration or similar- but that's largely because my IMAP clients are in a farm of servers. The biggest challenge in doing this is atomic replacements. Right now I go through the following contortions: 1) APPEND the message. It contains the md5sum of the previous configuration message. 2) make sure there's exactly one message in there- if not, delete our current new message and start back at step #1 3) delete the old message. 4) expunge. If I crash after #1 but before #3, I can detect this by loading both messages- one will contain the md5sum of the other. If two clients both APPEND at the same time, we'll have #2 to sort that out- crashes between two appends, and #2, we can detect because two messages will have md5sums of another - and we can delete them both. Having a single: REPLACE command would make that easier. Having it everywhere would be better. Of course, if keywords were everywhere, I could simply add/remove keywords to a dummy message. Keywords aren't everywhere (however), and so this isn't an option. -- Internet Connection High Quality Web Hosting http://www.internetconnection.net/
