[NNTP] Re: MODE READER

Russ Allbery rra at stanford.edu
Wed Nov 10 10:25:12 PST 2004


Clive D W Feather <clive at demon.net> writes:
> Russ Allbery said:

>> QUIT is indeed also a general command.  (In practice, STAT probably is
>> too, but we didn't put it into that category for some reason that I
>> don't recall off-hand and which isn't important enough to revisit.)

> Um, STAT *is* in that list. Did you mean something else?

No, sorry, I just missed it and confused myself.

>>> (4) Switching. Initially reader commands cannot be used; they get a
>>> 401 response. Once MODE READER has been issued, transit commands
>>> cannot be used; they get a 502 response. MODE READER destroys all
>>> server state apart from the switch and therefore LIST EXTENSIONS
>>> output can change in this one case.

>>> Is this correct, in people's opinion?

> But is it what we think happens in real life?

Mostly.  INN's nnrpd can be configured to support IHAVE, however.
(Although I'm not particularly happy with that IHAVE support; it's
intended for the use of broken Microsoft clients that insist on feeding
with IHAVE, so it still adds trace headers.  It may be better to ignore it
as something outside of the standard protocol.  It's not enabled by
default.)

> My intention is to rewrite various parts of the document to talk about
> "reader" and "transit" connections. Some commands will be tagged as
> reader commands, which need not be supported on transit connections, and
> transit commands, which need not be supported on reader
> connections. Then we talk about using different ports, heuristics,
> both-transit-and-reader servers, and, finally, MODE READER (if we don't
> kill it). I already have a fair amount sketched out in my head.

> How we advertise this becomes the least of our concerns. Really.

Oh, okay, I see what you're getting at.

As long as readers can still advertise transit commands and vice versa, I
think that sounds great.  And that would let us introduce 433 in a more
comprehensive fashion and maybe push people a little towards using that in
situations where MODE READER is currently used.

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



More information about the ietf-nntp mailing list