e-mail sync options?

Bruce Dawson jbd at codemeta.com
Mon Aug 30 14:08:53 EDT 2010


 On 08/30/2010 11:14 AM, Joshua Judson Rosen wrote:
> I could just write a suite of code to implement this myself,
> but maybe someone can save me the effort by pointing me to a canned
> solution :)

A many-to-many sync is extremely resource intensive. And the optimizations
for doing it with fewer resources are mathematically difficult. I'd like to
test it if you think you've come up with code to do it! (I've done this
in the
past with sync'ing databases, and gave up on all but the one-to-many
scenario.)

I'm not satisfied there are any (commercial or otherwise) products that
do the generalized many-to-many sync correctly. (Including tracking
deleted and attribute records.)
> It really seems like IMAP is the only thing available that's
> at all like a `portable, ubiquitous mail-handling API'..., so
> maybe what I want is to just use OfflineIMAP in IMAP-to-IMAP mode?
> If so, what sync-*topology* should I be using? Do I have to
> use a star topology, where everything syncs with each other
> through a central node? Or can should the IMAP UUID support
> be sufficient for me to do an N-way distributed sync with
> data flowing every which way?

I would recommend using a "master" node that all the satellites sync to.
I believe
an N-Way distributed sync will eventually end up in a mess that you
won't be able
to unravel (other than just declaring one the winner and dup'ing that to
all the other
potential "master" nodes). I say this given the unreliability of network
connections,
software releases, humans, and the world in general (e.g. practical
experience).

--Bruce


More information about the gnhlug-discuss mailing list