ietf-nntp Response from ARTICLE

[Ned Freed]

> > Thus for mail, recipient unavailable is a 4xx issue while nonexistent
> > domain is a 5xx one. On this basis, 5xx is the right group for permission
> > denied.

> Well No! It is the right return for "permission denied because you are
> unwelcome to speak to this server", but it seems an overkill for
> "permission denied to read this particular article, but please feel free
> to ask for other articles (for which permission will likely be granted)".

You're confusing the nature of the error with the scope within which the error
applies. The two are largely unrelated and only the former is reflected in the
first digit of a response.

> That was precisely the situation that started this thread. Seemingly it
> was a net.* article (still not clear to my why NEWNEWS offered it to me,
> because I had not asked for any net.*, but that is another issue).
> Seemingly, I am not trusted to read net.*. But my software (a much hacked
> version of nntpxfer) decided any 5xx return was fatal (I have fixed it
> now, of course).

> So I would have thought some 4xx response would be more appropriate in
> this situation.

A 4xx response is only appropriate if the problem is something that will
eventually be resolved, so that were the client to repeat the same set of
actions they would work at some future time.

FWIW, we sometimes see similar misunderstandings in email. For example, I've
seen clients which, when a RCPT TO (mail recipient) is rejected with a 5xx
error, fail the entire message even though there are other valid recipients.
But this is broken client behavior; it is not justification for asking servers
to issue a 4xx error in such cases when they really mean 5xx.

				Ned