"Harald" == Harald Hanche-Olsen <[EMAIL PROTECTED]> writes:
Harald> Careful, or a crash at the wrong moment might cause a mail
Harald> to be moved but not forwarded.
Of course. I intentionally left out the part about running an
occasional cron job (or some such) to check for orphans. We need to
leave some exercises for the reader.
Harald> I think the race condition
Harald> you are refering to is more benign: It might cause a
Harald> message to be forwarded more than once (extremely
Harald> unlikely), or a qmail-inject or mv to fail because the
Harald> message was yanked under the script's feet (more likely,
Harald> since the user may be reading his mail at the same time
Harald> the cron job is running).
True.
Harald> In fact, even moving the
Harald> message first does not get rid of this latter sort of race
Harald> condition, since there is a time interval between find
Harald> discovering the file and the mv command running (or
Harald> between the user's MUA seeing the file and trying to move
Harald> it).
Ah, true again.
Harald> If the MUA is not able to cope with the possibility
Harald> of a message disappearing like that, then the user must
Harald> precede the reading of his mail by first moving the
Harald> messages using a more fault tolerant program. My simple
Harald> script could also need some work in that area. I never
Harald> intended it to be a complete solution, just a reasonable
Harald> starting point.
Agreed. The user of the script should be aware of the race conditions
in any case.
j.
--
Jay Soffian <[EMAIL PROTECTED]> UNIX Systems Administrator
404.572.1941 Cox Interactive Media
