[NNTP] LIST EXTENSIONS (again)

[Russ Allbery]

Mark Crispin <mrc at CAC.Washington.EDU> writes:

> Checking for explicit codes has a long history of being problematic in
> SMTP and FTP.  NNTP has a dismal history as well (need I remind you of
> 380 vs 480)?

Well, I don't know about the other folks on this list, but I have to admit
that I'm not feeling particularly guilty, nor think the NNTP community as
a whole should be blamed, for status codes that were never documented
anywhere, that no one else had ever heard of, and that have not been
enountered by anyone else that I've talked to other than you.

I don't know what gave you 380, but it was broken by any documentation the
NNTP community has ever produced.  Either it was broken as written, or
it's some highly obscure or experimental server from before AUTHINFO, or
something else strange was going on there.  I certainly wouldn't blame you
if you just gave up entirely on any server that returned a 380 status
code.  People write hideously broken software; I don't think the IMAP
community is responsible for, say, some of the things that Courier does.

That being said, it has been documented in NNTP for quite some time that
clients should respond to unknown codes based on the first digit, so
clearly this is indeed a design goal and status codes should be chosen to
enable that behavior.  I think we're doing fairly well so far with this.
Note that such things as reactive authentication or reactive TLS
negotiation aren't going to work with that paradigm, but I think we're
moving away from that towards LIST EXTENSIONS advertisements, so that
isn't really a problem.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>