ietf-nntp Backfill: not again.

[Alan Barrett]

> If there is a third choice, that of telling servers to do new stuff that
> is well understood, not too onerous, but not widely implemented, then I
> would go for the third choice.

Hmm.  I wonder if I have just come up with that third choice.

A few people have suggested things like "If article 14 arrives before
article 13, just have the slave hide article 14 from the newsreader
clients until such time as article 13 arrives".  The problem with that
is that article 13 might never arrive, because it might have been
canceled or expired.

But what if we had a way for the master to tell the slave "Article 13
will never arrive"?  It would need a protocol extension and/or a special
article format (for example, something that looks like an article with
an Xref line but almost nothing else), and it would need the slave to
keep track of the difference between {articles that have not yet arrived
but are expected to arrive later} and {articles that were canceled or
expired either before or after arrival}.

Given that, the slave could hide article 14 from the users until such
time as article 13 either arrives or is reported as will-never-arrive.

--apb (Alan Barrett)