On 11 Aug 2006, at 11:15, Ken Murchison wrote:
Wesley Craig wrote:
Yeah, I have a mixed test cluster up, that I'm using to fix the
various xfer problems. Pass it along, and I'll test it.
Here it is. I removed all of legacy parameter crap. My new
'theory' is to store and pass around ACLs including the legacy
macros (in the same manner as the GETACL response). The server
then does the 'right thing' when cresting the rights bitmask.
Ken, I've tested this in our mixed 2.2/2.3 backend environment. It
seems to behave correctly for all 2.2 <-> 2.3 xfers. I'm not running
any 2.2 frontends, but it's clear to me that the code will function
correctly in that case as well. This is a very elegant solution.
Thank you for it!
Paul recently submitted:
https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=2866
which fixes:
1) reconstruct removes already expunged messages, defeating delayed
expunge
2) reconstruct doesn't tell the "sync" service to replicate its changes
7) reconstruct should notice and correct mailbox ID conflicts
We still have these open issues:
3) xfer doesn't replicate expunged messages
( https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=2870 ):
4) xfer onto a replicating backend causes sync_client to exit
5) xfer off a replicating backend doesn't remove data from the replica
8) renaming users causes sync_client to exit
And the biggie ( https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=2871 ):
6) removing folders doesn't benefit from delayed expunge
What will it take to get a workable solution to this problem included
in production releases?
:wes
