[ietf-nntp] OVER message-id
Andrew - Supernews
andrew at supernews.net
Thu Dec 18 17:30:03 PST 2003
>>>>> "Ken" == Ken Murchison <ken at oceana.com> writes:
>> At some point we put "OVER message-id" back into the protocol.
>> Andrew from Supernews is objecting to this, saying that it's
>> impractical in "every major server design".
Ken> Just out of curiosity, if its impractical in "every major server
Ken> design", shouldn't more than one person be voicing displeasure?
Ken> I'm sure Russ would've shot this down earlier if INN would have
Ken> issues with it.
INN is only one of several server designs these days, and I haven't
seen anyone actively involved in Diablo development participating here
any time recently. Our server differs from both.
I objected specifically this is one of those cases that _seems_ easy if
you look at it only from the point of view of INN or any "naive" server
implementation, without considering the kinds of things that large-scale
systems need. Your comment here:
Ken> If a server can fetch an entire article by msgid, is it that far
Ken> of a reach to get the overview data?
is typical of that, because it _is_ a reach - a distributed server
design (such as, for instance, Diablo) separates the article store
(accessed by message-id) from all the per-group and overview
information (accessed by group/article number). In such a system it is
highly desirable to be able to store articles without knowing their
assigned article numbers (and hence not being able to store the Xref).
>> He suggests, instead, an extension to NEWNEWS that would return
>> overview fields along with the message-ids.
I suggested that as being the correct fix for a proposed class of
clients that used multiple OVER <id> commands in order to filter
NEWNEWS results to decide what to download. Note that _no such clients
currently exist_, because XOVER <id> (and even XHDR <id>) is not
commonly supported; clients that use NEWNEWS and which to pre-filter
articles currently do so using HEAD <id> commands.
>> For the same technical reasons, he suggests that:
>> HDR Xref <message-id>
>> HDR :lines <message-id>
>> may be problematical.
Ken> Same as above, client authors should determine if its useful.
The current evidence here is that clients do not need HDR (anything) <id>,
because we've had it disabled for the past five years and nobody's ever
commented on it.
--
Andrew, Supernews
http://www.supernews.com
More information about the ietf-nntp
mailing list