ietf-nntp PAT

[Russ Allbery]

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

>> If INN refuses to do the PAT search for some reason (such as the header
>> not being in overview and not being willing to search articles on
>> disk), PAT returns an error code *not* an empty successful response.

> If we support this behaviour in PAT, it needs to be explicitly
> documented.

Yes, it absolutely needs to be explicitly documented, since some servers
have returned the empty successful result instead, which is bad.

> It should also have an error code that is *not* the usual 502 "not
> permitted". I'm not sure if this is a suitable case for 503 (I'm looking
> at 502 and 503 separately), or if it should be something new, in which
> case I would go for 521 but others might think 505 is better.

I don't have any immediate opinion here.

> While I'm at it:

>> See RFC-1036 for a
>> list of valid header lines.

> This implies that a server should reject any header line not in RFC
> 1036.  Is this the intention (I hope not) ? Or should PAT try to match
> any header that it's given (simple and useful) ? Or should the *server*
> decide which headers are allowed (see above) ?

The server should decide and return an error code if it doesn't like the
header.  (I'd hope that most implementations would support matching on any
header, but sometimes there are performance reasons not to do so.)

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