My maildir reader (<URL:http://multivac.cwru.edu./nnmaildir/>)
currently sorts messages by their modtimes; this is intended to show
the messages in order of delivery. In case a message is edited after
delivery, it'd be nice to sort by filenames instead (not by simple
lexicographic sorting; I'd have to parse the filenames). This means
the timestamps in the filenames should match the actual delivery times
as closely as possible.
<URL:http://cr.yp.to/proto/maildir.html> says:
# If you're delivering several messages from one process, use
# starttime.pid_count.host, where starttime is the time that your
# process started, and count is the number of messages you've
# delivered.
In my case, the program doing deliveries is a long-running UI program.
It's likely there will be long intervals between deliveries from the
same process. Is there any reason I shouldn't update the timestamp
for each delivery, so the filename-sorting works? In case there are
multiple deliveries close to each other, I'll still use _count, but I
could reset it each time the timestamp changes.
Interoperation with delivery processes other than qmail-local (or
other single-delivery processes using the same filename scheme) and
nnmaildir itself will probably be rare enough that I don't care too
much if the sorting is less than useful in those cases.
paul
