On Tue, 24 Sep 2002 17:34:09 MDT, Ben Parker <[log in to unmask]>  said:

> The theory of SMTP error codes (see RFC821) requires this of any/all mail
> transfer agent programs, including LSMTP.  5xx errors are classified as
> 'permanent' meaning retry will most likely not succeed.  The descriptive
> text after the 554 may offer slightly more of a clue.

Yes - but RFC821 *also* says that a 4xx error should be returned for
conditions that will probably succeed if retried later, without any user
intervention.  So "system load exceeded" should cause a 4xx reply, because
there's a good chance that if the sender system retries in 30 mins, the
load will have dropped.  On the other hand, "user over quota" is specifically
mentioned as needing a 5xx code - although it's *possible* that the recipient
will be under quota at the next retry, it won't happen unless a human
intervenes and cleans out some mail.  (I have seen systems that *properly*
return '4xx User Unknown' errors if they were unable to get a read lock
on a user database and waiting is impractical because you're not supposed
to delay the '250 OK' *too* long...

Given that Hotmail usually reports *other* 5xx errors with a fairly good
explanatory text like "over quota" or "deactivated user" and so on, the
current spate of '5xx Transaction Failed' sounds like plain crack-induced
bogosity - especially in conjunction with all the connection reset and
refused errors I'm seeing.

Today's Sendmail tip:  If you're running 8.12 or later, you can use the
"queue groups" feature to push all the outbound Hotmail traffic into its own
queue directory, so it doesn't clog things up for functional systems...