So Ken had to head home today, and I've only really got tomorrow to work on things.
The tree is looking pretty good. Here's the major headline features: * CalDAV/CardDAV * mboxevents * per message annotations * autocreate * sync-protocol-based XFER (not yet ported from Ken's 2.4 branch, I'm working on that tomorrow) And these major internal things: * twoskip database format * new mailboxes.db DLIST based format, including uniqueid, mtime and tombstone records) * index namelock release (this sounds boring, but it means that deletes and cleanup work with long-running connections. It's a big deal) * improved SELECTed index memory efficiency (48 bytes per message down from 128 bytes) * reliable idled * reliable quota -f * extended quota types (num folders, num messages, annotation use) as well as just STORAGE. Replaces "user_mailbox_limit" hack with a per-user field. * proper standards-compliant SPECIAL-USE, always on. * unixheirarchysep is the default Our goal is to release a pretty-much feature complete alpha this week, and ask everyone to test it soon. We would like to have 2.5 ready for inclusion in Debian Jessie, hence the time pressure. What didn't make the cut? * Xapian indexing (administration isn't really easy with current tooling, and it requires a forked library) * Conversations (we have plans to change the internals significantly, and don't want to have to support an existing implementation, plus it's messy code) * per-user modseq/uidvalidity (but look out for it soon, because it will be needed for even-more-efficient Cal/CardDAV sync) * new mailboxes.db key format that makes LIST more efficient and removes the need for improved_mboxlist_sort and hacks. We really need more documentation as well. I've been awfully lax at documenting my work. Bron. -- Bron Gondwana br...@fastmail.fm
