kaih at khms.westfalen.de
Fri Aug 12 13:40:00 PDT 2005
clive at demon.net (Clive D.W. Feather) wrote on 11.08.05 in <20050811082704.GE51754 at finch-staff-1.thus.net>:
> Finally, think how it actually would go into the wording. To remind you,
> this is what the text currently says:
> The second type of key is
> composed of a newsgroup name and an article number within that
> Article numbers MUST lie between 1 and 4,294,967,295 inclusive.
> [In the following I'll use the "2^N-1" notation for clarity.]
How about something like:
Article numbers MUST be larger or equal to 1.
Note 1: Some current server and client implementations limit article
numbers to 2^31-1.
Note 2: It is expected that some servers will exceed article numbers of
2^32 a few years after publishing this spec.
That is, do it as advice to implementers, not as a MUST or SHOULD. "This
is what you will actually encounter."
NB. It so happens that I know of another implementation that seems to use
global article numbers, though it isn't anywhere near the number limit as
it doesn't actually take a feed. It is a NNTP access module to a web board
(and the only name for the software I know is "Webboard"), which
incidentally is quite buggy as I found out when using my normal NNTP pull
script on it :-/ I think one of the half a dozen or so problems was
actually my bug (pipelining too early).
The installation I know about is on bar.baen.com. (For one, it currently
uses message ids lacking any domain part. Needed to patch my local INN to
make it accept the things. It's an excellent mispropagation prevention aid
More information about the ietf-nntp