Hello,

There was a conversation in May about a replication strategy.

It seems Cyrus doesn't have it right yet: http://cyrusimap.web.cmu.edu/imapd/install-replication.html

Any news and roadmap on this topic?

Thanks,

Michael.

On Fri, Nov 02, 2007 at 03:48:41PM +0200, Timo Sirainen wrote:

Actually it caches everything that is used. http://wiki.dovecot.org/IndexFiles and http://wiki.dovecot.org/Design/Indexes/Cache tell about the exact rules.

I've read those docs, indeed, and I've run a test against 1.0.5. I see ~0.25s response time for a SEARCH on eg msg-id but ~15-35s on X-something. After that 1st run, SEARCH on same X-* takes ~same (~10kmsg mailbox). So I guessed index cache is done on the fly for any SEARCHd for item, but in memory only, since checking index files after test didn't show such X-something. But re-checking now, I realize I must have checked index files too early: my custom X-something are right there into index.cache. So I think it might already be smart enough for our purpose :)

thanks

-- paolo

GPG/PGP id:0x1D5A11A4 - 04FC 8EB9 51A1 5158 1425 BC12 EA57 3382 1D5A 11A4