[NNTP] Re: [ietf-nntp] draft-ietf-nntpext-streaming-01.txt

[Russ Allbery]

Ken Murchison <ken at oceana.com> writes:

> I still think that "not wanted" (a.k.a. "already have this msgid") is
> still valid for TAKETHIS.  This could occur because a client tries to
> send an article immediately with TAKETHIS (w/o having done CHECK first)
> or because of the race condition between running CHECK and TAKETHIS
> (e.g. another process accepts the particular message in the
> interval). Having said that, 439 would suffice for this case, but I look
> at that as "not accepting articles for this group" rather than "already
> have this msgid".

> I'll defer to the consensus on whether to document 438 for TAKETHIS.

Well, I've never seen an implementation that did that; all the
implementations I'm aware of send 439 for that case.  There is no
distinction drawn in the protocol between duplicate messages and messages
rejected for any other reason, just between messages rejected before
they're sent and messages rejected after they're sent.  (They could be
rejected before they're sent for other reasons than being duplicates --
for example, INN can be configured to query an external filter for each
message ID sent via CHECK or IHAVE and allow rejections in advance based
on some external knowledge.)  Reports and statistics are based on that
distinction too (for example, innfeed's statistics report the counts of
439 responses vs. 438 responses, intending them to be used in the above
sense).

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