[NNTP] Fwd: Gen-art review of draft-ietf-nntpext-streaming-05

[Ade Lovett]

Russ Allbery wrote:
> Could you look over the text I posted earlier this morning and see if it
> would work for you?  I think there are a few things we can usefully say.
> 

I've read it over and I have to admit that I'm siding with Ken on this
one as being more of an implementation issue than anything.

We're already saying (in nntpext-streaming sec.2.4.2):

	"Note however, that the responses to CHECK are advisory;
	 the server MUST NOT rely on the client to behave as
	 requested by these responses."

To me, the strong MUST NOT covers everything we'd want to say at a
protocol level.  As to exactly how it should be dealt with, that's
definitely treading into implementation specifics.

The most likely approach taken by a server in such a runaway case would
be to simply terminate the TCP session, doing anything more gracefully
is unlikely to work, and would be syntactic sugar anyway.

Client behavior in the case of the TCP session going away is already
covered elsewhere, and it would presumably try to reconnect.  As to what
the server does at this point is again an implementation detail, eg:

	"502 bad runaway client, no cookie, contact news at example.com"

-aDe