On Tue, Jul 19, 2011 at 05:09:01PM +1000, Greg Banks wrote: > G'day, > > My work to implement RFC5257 and RFC5464 is nearly ready to merge to > master. It needs a couple of polishing touches, but most of the > work is done and I would appreciate some expert review (normally I > would just ask Bron as he's been tracking this, but he's on > holidays).
Bron will grab a copy of it as well - and have a poke through, but not sure how much time I will have. Particularly since I've been dragged into arguments about missing parts of IMAP again. I appear to have signed up to write specs for the following: 1) DELIVER - take a message via APPEND or from the spool and shove it to sendmail or moral equivalent for delivery - possibly add a couple of options to control it, but don't make it so difficult that nobody will implement it. 2) MOVE - someone already has a half-done spec for this, but they want it to be atomic, which makes it complex to implement, so nobody will. Write a spec for simple. Worst case, extend the current spec with a NO_ATOMIC_MOVE to match the NO_ATOMIC_RENAME 3) DIGEST - specify a way to fetch a digest of a message (or maybe even part). Both to get at the SHA1 we already store, and optionally to allow digests of decoded parts! Gotta think a bit about this one. CPU is getting cheap enough that on-the-fly digests aren't unreasonable. 4) GUID - in our case it's just the SHA1, but gmail have their X-GM-MSGID and I'm sure other servers have other ways to uniquely identify a message over moves, even by other clients. So you can keep the raw data cached locally in the presence of other clients. Bron.
