Mail Servers
John Chambers
jc at trillian.mit.edu
Sun Sep 21 16:05:14 EDT 2003
Rich Braun writes:
| dsr <dsr at tao.merseine.nu> wrote:
| >> ...some arguments from folks who have made the switch from mbox to
| >> maildirs.
|
| Aside from the performance and reliability issues you raised, are there
| software capabilities available in maildirs format that cannot be implemented
| in plain-text mbox format?
Well, one "capability" that I like is that I can easily augment any
email package with my own perl scripts to do useful things. Doing
this with any of the various "mbox" formats is difficult. First, it
takes extra time to write the routines to demux the messages and then
recombine them. And then, all too often you discover that you've done
something subtly wrong with the format that confuses some other
software that you're using. With each message in its own file, you
can just use the usual filesystem operations, and it's difficult to
get that wrong.
The trickiest part with mboxes is getting the locking right. It's
never documented, and if you don't do it right, you risk losing
incoming messages when you try to do something with your inbox.
One question I'd have, though: Are there any general guidelines (or
even - gasp - standards) for how to do things portably with maildirs?
I've seen different packages differ in things like how to indicate a
deleted message (without actually deleting it), how reply messages
are named, and how to indicate the current message. I suspect that
there aren't any actual standards for these things, so my own code
can't be completely portable. If there are standards or even common
conventions, I'd like to know about them. If there are important
warnings and gotchas, I'd like to know about those, too.
OTOH, I have so many email-munging perl scripts now that I may soon
be independent of any actual mail reader package. I'm pretty close to
duplicating everything that I use in my own code. Maybe I should just
finish the job. .-)
--
O
<:#/> John Chambers
+ <jc at trillian.mit.edu>
/ \ <jmchambers at rcn.com>
More information about the Discuss
mailing list