ietf-smtp
[Top] [All Lists]

Re: handling extraneous 500 level SMTP error codes

1998-04-10 18:33:16
just going through a couple of bounce messages that I have
received back from mailing lists, and came across a couple
of instances that cause me to question my responses.

The first is responding to a MAIL FROM with a
551 <address(_at_)domain(_dot_)com>... <ISPs NAME> ERROR 1 - call <#> for 
help.

The second is responding to a RCPT TO with a
554 Could not read response from aliasd.

Now, in both cases, I could not find (in 821 or 1123) the response
codes listed for the commands.

While I try and use documented codes in the code I write, I've never bought
into the theory that servers which issue undocumented but structurally legal
response codes are broken. RFC821 makes it quite clear that the response code
set is intended to be extensible. It also makes it clear that interpretation of
specific codes really isn't a good thing.

Matter of fact, user not local for
MAIL FROM does not really make sense unless they have an anti-posting
filter or something out in place.  Seeing as this is list traffic,
it really didnt make sense.

People implement rejections at the MAIL FROM for all sorts of reasons. While I
tend to regard such things as fairly problematic, the fact remains that doing
this sort of thing is fairly popular and people believe they are well served by
it. And I don't think clients have any business second guessing servers in this
regard. For all you know this server has had a problem with mail that had the
MAIL FROM address you are using and is perfectly justifid in blocking it. If
the server is doing the wrong thing the right outcome is to fix it, not try and
work around it.

The second case I can almost see treating this as I do thr 550
to 553 responses, which indicate failures with the particular address,
but not sufficient to abort the connection (like mailbox full, etc.).

Actually this one's a bigger problem for me. This sounds like a case where
a 4xx code should have been issued (or the address accepted and alias
expansion deferred) rather than a 5xx code. But even in a case like this
I think it is better to take the code at face value and bounce the message.

So, any input there on how these 500 level response should be
handled?  current usage of any of these responses to these commands?

I'd handle them by bouncing the message.

                                Ned

<Prev in Thread] Current Thread [Next in Thread>