On Fri, 26 Sep 2003, Marcel Crasmaru wrote:
>--- Rob Siemborski <[EMAIL PROTECTED]> wrote:
>> C2, of course, wins in this case if it just gets and holds the mailbox
>> lock first.  In fact, in this case C1 has *no* chance of ever deleting the
>> message if it loses the first race.
>Yes,  but it is a different situation. A lock in the protocol gives
>explicit concurrency control.
>Anyway, IMAP is not SQL :) and maybe explicit locking is not
>a wise idea right now. IMHO,  for an IMAP server developer explicit
>locking may be a fancy feature, but if I were an IMAP client
>developer for small devices I would like to have an explicit
>concurrency control.

You get the best concurrency in a system that is designed so that no
locking is needed for writing nor reading (for instance, with mail
delivery to a Maildir).

If IMAP would introduce an explicit locking mechanism, I would have to
claim it was a step back in terms of the protocol's evolution. :-) Rather
define (or discover) semantics that allow one to avoid locking altogether.

Andy

-- 
Andreas Aardal Hanssen
http://www.bincimap.org/

Reply via email to