Re: [Nmh-workers] Updates to nmh

[Ken Hornstein]

>>While I haven't encountered any problems with nmh's SMTP code, if you know
>>of problems with it I'll be glad to fix them.
>
>Just to pick a random example, if you get a 4xx response back from the
>server you're supposed to retry later, and I bet we don't do that :-)
>Another one: I don't think we do MX lookups on the server name given in
>mts.conf, which is also a MUST from RFC1123.

Those are both requirements for SMTP MTA's, that is true.  But since
nmh isn't doing final delivery, I don't think those things apply.

People smarter than I can speak up, but I think nmh falls more in the
category of a mail _user_ agent, and thus RFC 4409 is more in line with
what nmh should be doing.  But specifically:

- Should we retry on an error?  Well, yeah, probably.  That would be
  difficult, I'd think (I believe you would want to back off and not simply
  retry right away).  Right now we simply notify the user about the failure.
- Since we're not doing final delivery, the server listed in mts.conf file
  is not a _domain_ (part of an email address) that we're connecting to, but
  the name of server that runs an SMTP service we can use for message
  submission.  As such, MX records are not appropriate.

Both of these things are things that plenty of other MUAs do.  That doesn't
necessarily mean that nmh should be doing them as well, but there is plenty
of precedent.

--Ken