[ietf-nntp] :bytes metadata

[Russ Allbery]

Ken Murchison <ken at oceana.com> writes:
> Clive D.W. Feather wrote:

>> I'm not sure what to do about (1). I can think of four possible fixes:
>> (A) Say that the value applies to the current session and note that, while
>>     it SHOULD be constant across sessions, there are cases when it isn't.
>> (B) Require :bytes to be a maximum in cluster cases; is this practical?
>> (C) Allow it to be one, two, or three values:
>>         actual
>>         actual,max
>>         min,actual,max
>> (D) Have another metadata item reporting the potential problems with the
>>     value.

> I'd say (A).  A cluster is a specific implementation and we shouldn't
> munge the protocol to account for idiosyncrasies specific to them.

I agree with Ken.  I think that's the simplest fix, since I don't think
that a maxium can be predicted.  (C) and (D) are too much complexity for
the nature of the problem, IMO.

The important thing here is to warn that clients cannot rely on the bytes
count for more than an estimate, which I seem to recall that we already
do.  It's mostly useful for filtering and size estimates, not for any sort
of core protocol purpose, so a bit of minor deviation isn't a huge deal.

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