On Sat, Oct 09, 2010 at 07:30:36AM -0400, Wesley Craig wrote: > On 09 Oct 2010, at 00:04, Michael D. Sofka wrote: > > Is the old format updated on xfer? > > It is updated on xfer -- when the mtimes of the message files are set to the INTERNALDATE. Yes, file mtimes get set to INTERNALDATE as much as possible (and vice versa when reconstructing with a missing index - hence the reason for doing it) XFER itself is a super-dumb protocol (I should know, I just rewrote it to work correctly with namelocking in 2.4, and I have one more patch outstanding to discuss with Ken on Monday...) All it does is collects up ALL the message files and meta files and copies them across the wire as-is. No consistency checking, no upgrades. They will be upgraded at the remote end when the mailbox is next opened. I'm planning to replace it with something better... ... and when I say replace, I probably mean "co-locate". To the point that an incoming XFER will cause an automatic upgrade at the point of the XFER, and an outbound XFER will probably actually generate a 2.2.x compatible index file. I don't see the need to support older than that. Cyrus will also advertise a new capability for replication-based transfers which will have much more integrity checking, and will be used if both ends are recent enough. Bron. ---- Cyrus Home Page: http://www.cyrusimap.org/ List Archives/Info: http://lists.andrew.cmu.edu/pipermail/info-cyrus/
