Kevin Sullivan <ksulliva@psc.edu> writes:

> --On 12/02/02 01:15:42 -0500 Eric Gillespie wrote:
> > Unless you have to support users who *need* mbox, go with
> > maildir.  Delivery to a maildir is atomic, while mbox is
> > guaranteed to give you a corrupt mailbox once in a while.

> Not true.  While the mbox format is poorly designed, corruption
> will only be a problem if you use mail over NFS or if your mail
> clients and delivery agent don't agree on a locking method.
> The big problem with mbox is that you cannot have multiple
> processes with simultaneous write access.

Sorry, you are wrong.  Forget NFS and assume ideal mbox
conditions.  When any failure occurs (machine crash, power loss,
filesystem full), a message ends up truncated.  The corruption
comes in if a message is truncated in the middle of the line.
Then, when the failure condition is corrected, the message is
delivered again.  Delivery to a maildir, on the other hand, is
atomic.  Maildir is immune to the reliability problems of mbox.

I won't reply to more FUD, intentional or otherwise.  These
issues are heavily covered all over the Internet already.

> Maildir is not supported by many mail clients and it suffers
> from the standard one-file-per-message, no-header-database
> problems (same as MH format).  I don't care for it, but your
> mailbox format will be driven by both your programs and your
> needs; it may work for you.

Sure, but the confusion comes from people assuming the user's
*storage* format must match the *delivery* format.  For storage,
mbox isn't so bad after all.  As a matter of fact, i archive my
mail with mbox (live mail is in mh folders).  But using anything
other than maildir for the *delivery* format is guaranteeing
failure.

In many (most?) cases, such as the one the original poster
described, the format is irrelevant completely.  Users will read
mail with IMAP or web mail.  Even on systems where users read
mail locally, the popular Unix MUAs support maildir these days,
and programs to incorporate mail from a maildir maildrop into an
mbox for reading by elm or whatever are trivial and do exist.

> I'd caution against the Courier IMAP server; it disobeys the
> IMAP RFCs in subtle ways, and the author loudly claims that the
> problems are in the RFCs not in his program.  But it works with
> most IMAP clients, and many people seem happy with it.

I have heard of this and consider it unfortunate, but unless
you're willing to go with Cyrus and its custom storage back-end,
Courier is the most viable option :(.  I can say that i have
never had any users experience problems due to Courier IMAP.

--  
Eric Gillespie <*> epg@pretzelnet.org

Build a fire for a man, and he'll be warm for a day.  Set a man on
fire, and he'll be warm for the rest of his life. -Terry Pratchett