ietf-nntp Article numbers and "Replaces"

[Charles Lindsey]

There is a statement in 10.1 in draft-ietf-nntpext-base-04.txt that:

"The server MUST ensure that article numbers are issued in order of
arrival timestamp;"

OTOH, there is an intention (see draft-ietf-usefor-article-01.txt) to
introduce a Replaces: command, whose effect is the same as the
Supersedes: command except that it is intended that a replaced article
should not normally be offered to the reader for viewing (it being
presumed he has already read the earlier version). This is intended to be
used for regularly posted articles such as FAQs in cases where the
reposted article is identical to the earlier one (only reposted because
the old one had expired) or where the changes are considered too trivial
(fixes of typos, etc.) for it to be drawn to the attention of every
reader. This is considered to be a useful new feature by the USEFOR
working group.

The simplest and most obvious way to implement this new feature is for
servers to store the new version under the same article number as the old
one. This means that existing newsreaders which rely on a conventional
.newsrc file will automatically get the benefit of the new feature without
the necessity of any upgrade. It is, however, a technical breach of the
statement I quote above, which suggests that that statement ought perhaps
to be qualified in some way - certainly not a "MUST" as at present.

The only problem with that simple implementation of the feature is that it
will leave the low water mark for the group permanently at a rather low
number, and implementations that report the number of articles in the group
as the difference between the high and low water marks (see 10.1.1.1 -
GROUP command) will be highly in error because the sequence of article
numbers in the group will become extremely sparse.

Now it has been suggested that this will break newsreaders that, for
example, use article numbers as indices into an array bounded by the
reported water marks. Is this in fact a regularly used implementation?
Certainly the nntp draft gives no guarantee that the space will not be
sparse, and indeed it suggests (10.1.1.1) that "reinstating" an article
with the same article number is a normal behaviour.

So my question is - should we encourage that sort of implementation by
qualifying the sentence quoted above, or should we discourage it by
leaving it alone (or even strengthening it)? And if we want to discourage
it, can anyone suggest a better way to implement it?

Note that this message is cross posted to both the USEFOR and NNTPEXT
mailing lists, and I have tried to set Reply-To accordingly (but my
gateway may not play). Please try to keep it in both lists.

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Email:     chl at clw.cs.man.ac.uk  Web:   http://www.cs.man.ac.uk/~chl
Voice/Fax: +44 161 437 4506      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9     Fingerprint: 73 6D C2 51 93 A0 01 E7  65 E8 64 7E 14 A4 AB A5