ietf-nntp DRAFT summary of IETF 49

[Russ Allbery]

Charles Lindsey <chl at clw.cs.man.ac.uk> writes:
> "Stan O. Barber" <sob at verio.net> writes:

>> If we specify HDR based on XHDR, then you are probably right. Today,
>> there is no HDR documented and I am concered that wildmats might creep
>> in there.

> Yes, that was the intention. Wildmats are fine for specifying newsgroups
> names, but are poor for more general applications. So we invent HDR just
> like XHDR and leave anything else for future extensions. I believe
> Andrew told us that XHDR is currently used far more than XPAT.

Yup, I definitely agree with this approach.  XHDR unfortunately consumes
more bandwidth to the client than XPAT, but it lets the client use its own
matching language and not be constrainted by wildmat.

>> I think this notion is a good one, frankly. However, I don't think
>> current implementations that involve multiple programs have the ability
>> to "switch back" once they have gone to "reader" mode.

> MODE READER would make far more sense if accompanied by MODE RELAYER to
> go the other way.

Only in the sense of meaningless logical consistency; basically no one in
the real world would ever use MODE RELAYER.

> AIUI, it was INN that invented MODE READER in the first place and got us
> into this mess.

Yes, I believe INN introduced MODE READER.  It's probably one of those
ideas that sounded good at the time.  I would have strongly preferred it
if two separate ports had been used since the beginning, but realistically
we have a *huge* installed base that isn't going to change any time in the
near or medium-term future.

> It might be sensible to say that you MUST use the MODE commands if you
> are doing streaming, but that they can be omitted otherwise. Would that
> work?

No.

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