[ietf-nntp] :bytes metadata

[Russ Allbery]

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

> I still have this outstanding.

>>> Actually this is a key point. What do clients do with the :bytes value?
>>> I believe there's a consensus for "SHOULD equal the canonical size",

> and I've amended the text accordingly.

>> I think Andrew is arguing for us making the following two statements:
>>> * :bytes MUST be greater than the canonical size minus 999.
>>> * :bytes MUST be less than the wire size plus 999.
>> Andrew, is that right?  Is 999 reasonable?
>> 
>> He has the most practical experience with the issues raised and with how
>> clients react to them in practice, so I'm inclined to go with his
>> expertise here.  The only change I may argue for is s/wire size/canonical
>> size/ in the second statement as well and just saying "tough" to servers
>> who don't have easy access to the canonical size, but I don't have a
>> strong feeling about that.

> I then listed a range of alternative conditions (e.g. using a multiplier
> rather than a constant).

> Do we have any more input?

One option is to not say that the :bytes metadata MUST be anything at all.
In other words, we could just leave it at a SHOULD, with the implication
that it's going to be somewhat unreliable in practice.

Unless we can reach consensus on a firm number, I think that's probably
the best approach, and it's the one that most accurately reflects the
current situation in deployed software.

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