[Dovecot] Replication via sneakernet

[David Bishop]

On Fri, Dec 6, 2013 at 10:28 PM, Nikolaus Rath <Nikolaus at rath.org> wrote:

> David Bishop <dovecot at dpe.lusars.net> writes:
> > Is there a way to get a running transaction log from a Dovecot instance
> to
> > synchronize (and I use that word very, very loosely) multiple instances
> > that don't share a network?
>

>
I think what you want is not synchronize the state of one mail server
> across multiple computers, but have mail transport from one system to
> another via sneakernet. In other words, you want UUCP. It might be
> ancient, but it should do exactly what you want.
>

I've looked at UUCP, and while it solves a related problem, it doesn't seem
like it solves my problem (which is why I originally asked about copying
files directly out of the spool). It is, I think, possible to force UUCP to
do what I want, but dsync seems an easier solution. (Sadly, it'll be a bit
before I have time to work up a testbed.)

As an example:
If Bob-from-maintenance is working his way up the line, he may be emailing
updates to the maintenance list (stuff like "this needs replacement but I
don't have the part" and "This problem has been sorted") before hopping a
tram to the next station.
At the next station (please note: different server with its own mail spool
pretending to be the same server), the local IMAP server should look (to
Bob, and Bob's phone) exactly the same (down to message IDs, because
otherwise, his phone is going to get confused about the locally-cached
email).
If an email someone sent in response to one of his emails catches up with
Bob ("I have the part and will sort it -- Ann-from-engineering" or "new
problem reported", for instance), it should show up as a new message, but
his original message should show up as having been read.

If I do this with dsync (and correctly understand dsync), this works the
way (I think) it should. If I do this with UUCP, every station is going to
have its own copy of each message, but not copies of read flags (and no way
of conveying that a message has been read, or, more importantly, deleted,
which becomes a problem when some mail recipients never go all the way to
the end of the line).