Re: #1416 Reinjection - an attempted summary, and a suggested resolution

[Russ Allbery <rra@xxxxxxxxxxxx>]

Charles Lindsey <chl@xxxxxxxxxxxxxxxx> writes:

> That is fine for well-written agents, but there are an awful lot of
> people out there writing posting agents wo think that news is just a
> minor variant of email, and I would not trust them to do it right. They
> need to ensure that the Injection-Date header is not added until they
> actually have a connection to an NNTP server open and ready to accept
> the injection.

Well, I think you've already proven by experiment that if they miss by a
day, it's not a big deal.  It's only a problem if they miss by three or
four, at which point we're mostly talking about off-line readers, which
are more sophisticated clients to start with.  I'm more confident in the
ability of people who write off-line readers to get this right than people
who write a simple normal client.

> As far as USEPRO is concerned, we have to allow either party to add the
> Injection-Date. As to what Best Practice might be, that may be more or
> USEAGE.

In the absence of any stronger arguments than "some clients are stupid"
(and right now people don't seem to be presenting any), I'm opposed to
breaking the existing identity model for existing clients written before
the introduction of Injection-Date.  I don't think that's the sort of
non-backward-compatible change that's okay for us to make just because
some clients get it wrong.  I think we can provide mechanisms for clients
that want to do the right thing to do it right, but potentially breaking
the identity model of currently well-behaved clients to work around
problems with others doesn't make me comfortable.

So I only agree with this statement if there is no Date header in the
proto-article.

-- 
Russ Allbery (rra@xxxxxxxxxxxx)             <http://www.eyrie.org/~eagle/>