[ietf-nntp] :bytes metadata

[Russ Allbery]

Russ Allbery <rra at stanford.edu> writes:
> Ken Murchison <ken at oceana.com> writes:

>> Do we really need this abstract bracketing of the value?  What's wrong
>> with stating that :bytes is an estimate which MUST NOT be less than the
>> canonical size?

> Significant portions of Usenet will ignore the standard because they
> won't re-engineer their software to have that number available (the work
> required in some cases is extensive) and in practice :bytes will be
> within a particular range of variation anyway.  Client authors will just
> curse the standard for not telling them how software really works.

That required some more explanation, and I'm responding too fast.  Sorry
about that.  (Fighting with issues at work.)

The problem is exactly that the actual article size will tend to be
slightly larger than the overview size, since the overview size is
generated by a different system (and therefore an additional Path header
may be added and Xref may be synthesized and added).  That's as I
understand it, at least.  I think the errors are likely to be on the side
of slightly underestimating the article size, so the above compromise
proposal doesn't help.

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