ietf-nntp Draft 17 pre-2

[Russ Allbery]

Ken Murchison <ken at oceana.com> writes:

> Well, yes and no.  If the client sends an initial response, eg:

> AUTHINFO SASL PLAIN AHRlc3QAdGVzdA==

> then the cat is already out of the bag.  If it doesn't (costing an extra
> round trip), eg:

> AUTHINFO SASL PLAIN

> the the server can return a 482 or something and the client should not
> send the response.

Hm.  Okay.

Well, I certainly don't have a problem with strongly recommending (a
SHOULD sort of thing) that clients check LIST EXTENSIONS first.  I don't
think it should be a MUST, though, since after all you can avoid doing so
without causing interoperability problems.

>> I think this is going to depend some on what extensions one is using.
>> In another five years, for example, I certainly won't expect clients to
>> send LIST EXTENSIONS before trying OVER.

> Why not?  It might be safe to assume that all servers implement OVER,
> but what's the harm in checking?

I guess it just feels like a pointless additional round trip to get
information that you could easily get in a single round trip anyway if you
actually need it.  But it's not a very expensive round trip, granted.

> Is there some technical reason why clients would/should/might not want
> to do option discovery,

Not that I can think of.

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