Re: road warrior trying to simplify a mail setup

[James Leifer <James.Leifer@inria.fr>]

[long reply...]

Greetings!

I face a similar problem of synchronising my mail on two machines that
are often offline.

I have been considering using nnmaildir to manage all mail (one
message per file).  If I understand Paul Jarc's description correctly,
this has the advantage of storing each message's mark in a separate
file (not together in a big .newsrc).  If so, then I could use unison
(which already synchronises bidirectionally my 600 MB home directory
in seconds) for the mails and marks.  Anyone tried this?

Unison is at: http://www.cis.upenn.edu/~bcpierce/unison/ .

Another interesting tool I noticed (but haven't had time to play with)
is offlineimap: http://gopher.quux.org:70/devel/offlineimap/

As far as I understand it, an ideal scenario looks like this:

 (1) Imagine two machines, A and B, that can connect to each other
     from time to time, each of which has a complete replica of my
     mail.

 (2) I make changes on machine A involving both the physical messages
     (moves between groups, creation of new gcc'd messages, deletes,
     etc.) and the marks (read, replied, etc.), with gnus keeping a
     log of what has changed.

 (3) Likewise, I make changes on machine B.

 (4) I connect A and B and some tool merges the two change logs and
     then (bidirectionally) copies and deletes files so that the
     replicas are again identical.  The merging might have to deal
     with conflicts that I need to resolve manually (e.g. on one
     replica I have deleted a message, on the other I have moved it to
     another group).

What would it involve to implement?

  (2) requires gnus to keep logs of message and mark changes. Can the
  agent do this?  (It might need to record things like "message with
  message-id x and hash y and file name z has moved from group c to
  group d".)

  (4) requires a tool to parse the log files and ask about conflicts.
  The actual copying could then be handled by unison.

Anyone tried this?

Regards,

--
James Leifer, INRIA Rocquencourt, France