article numbers (was Re: ietf-nntp Backfill)

[Chris Newman]

On Fri, 13 Dec 1996, Rich Salz wrote:
> Jon Ribbens wrote a very nice note.  Article numbers should be considered
> unique resources, never to be re-used unless there is a "change in epoch"
> which requires out-of-band coordination between client and server.

Oddly enough, IMAP4 had to deal with this same problem to support
disconnected operation.  The difference is that with all the legacy
mailstores out there that do stupid things, IMAP4 has the "change in
epoch" message in-band.

I recommend making the "strictly ascending in delivery order" rule *very*
explicit, as the IMAP community had some problems with interpretations of
RFC 1730, so the wording is tighter in RFC 2060.

I also recommend considering if you want to place an explicit 32-bit limit
on article numbers.  In the case of IMAP4, there were people planning on
using 64-bit or longer article numbers (IMAP4 UIDs) until the rules were
made clear.  I suspect clients may break if the article numbers get over
32 bits.

I consider it desirable to have IMAP UIDs and NNTP article numbers 
identical, since they serve the same purpose.