From: Russ Allbery (rra@stanford.edu)
Date: Thu Apr 29 2004 - 01:26:41 CDT
Charles Lindsey <chl@clerew.man.ac.uk> writes:
> Bruce Lilly <blilly@erols.com> writes:
>> Loops, multiple copies of an article (with different message-ids).
> What ARE you talking about? Where have those "different message-ids"
> suddenly sprung from?
It's an example of a problem caused by reinjection, and yes, it absolutely
does happen. It's one of the reasons why other news administrators are
sometimes somewhat unhappy with the way that Airmail handles message IDs,
for example.
It's not something that can be solved at the protocol level, but it's a
reason to strongly discourage reinjection and to take a very hard line
about not ever substituting one's own message ID if the provided article
already has one (which is what Airmail does). Unfortunately, some methods
of reinjection also lose the message ID.
There are a bunch of problems like this that Usenet would never encounter
if no one ever reinjected a message.
> However, under the present setup where the Date-header is used,
> reinjection does no harm because the Date-header never gets altered,
Reinjection *if done exactly correctly to a news server that handles it
properly* causes no problems *if all sites on Usenet handle history file
expiration properly*. Unfortunately, in practice, no matter what best
practices are or what we write in an RFC, not all of those conditions are
always true.
Robustness says that if you have a failure mode that can cause significant
damage (and reinjection loops can flood newsgroups, although thankfully
that's rare), you should have defense in depth and not rely on any one
mechanism to stop them. The best defense is, of course, not to engage in
the risky behavior in the first place.
> Therefore, the present draft had no need to mention any special
> precautions to be taken by injecting agents when reinjecting,
Encouraging reinjection is almost always a mistake. It's a last resort
that should only be used in very specific circumstances by people who know
exactly what they're doing.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>