Re: [sup-devel] [PATCH 2/2] Use nanosecond resolution of mtime for generating the unique id for each message

From: Rich Lane <rlane@club.cc.cmu.edu>
To: Mark Alexander <marka@pobox.com>
Cc: sup-devel <sup-devel@rubyforge.org>
Subject: Re: [sup-devel] [PATCH 2/2] Use nanosecond resolution of mtime for generating the unique id for each message
Date: Fri, 19 Mar 2010 17:17:50 -0400	[thread overview]
Message-ID: <1269032592-sup-2403@zyrg.net> (raw)
In-Reply-To: <1269030873-sup-4844@r61>

Excerpts from Mark Alexander's message of 2010-03-19 16:35:38 -0400:
> Excerpts from Michael Stapelberg's message of Fri Mar 19 15:47:53 -0400 2010:
> > Excerpts from Ben Walton's message of 2010-03-19 20:35:31 +0100:
> > > Isn't there a more portable solution?
> > There is: Rich mentioned that proper move/delete handling for maildirs also
> > solves the problem.
> 
> I must have missed that message.  What is the proposed solution?
> How does it solve the problem of out-of-order maildir IDs?

Notmuch implements this correctly. They essentially keep a sorted list
of mail filenames in the index and then iterate over that and the sorted
list of the filenames that are currently in the maildir to detect
inserts, deletes, and renames. 

I just rebased and pushed a multiple-locations branch I've had sitting
around to mainline. It's probably buggy. It's been a while since I
thought about this problem, but I think keeping track of multiple
locations (source/source info) for a single message (identified by
message-id) should make the maildir logic simpler.
_______________________________________________
Sup-devel mailing list
Sup-devel@rubyforge.org
http://rubyforge.org/mailman/listinfo/sup-devel