Re: gmane "Server Closed Connection" problem on Vista

[Steinar Bang <sb@dod.no>]

>>>>> Ted Zlatanov <tzz@lifelogs.com>:

> Strange.  It would be ideal if we could isolate the test case to be
> outside of Gnus.  Can you debug the NNTP interaction and find the
> command sequence that triggers the disconnect?  It should then be
> fairly simple to test it outside of Gnus (for instance, with Perl
> using Net::NNTP).

I have a wireshark sniff of a failed group open, and just looking at the
NNTP traffic doesn't yield any clues.,

The last thing that goes on there, is:

XOVER 9209-9209
224 9208-9208 fields follow
9208.Re: Languages again.Paul R <paul.r.ml@gmail.com>.Mon, 10 Nov 2008 12:28:31 +0100.<87zlk7q0ps.fsf@gmail.com>.<2AF10E75-89F8-49A9-BD07-F50E394C401F@gmail.com>.5048.21.Xref: news.gmane.org gmane.emacs.orgmode:9208.
.
224 9209-9209 fields follow
9209.Re: Languages again.Sebastian Rose <sebastian_rose@gmx.de>.Mon, 10 Nov 2008 13:07:33 +0100.<87ljvriy2i.fsf@kassiopeya.MSHEIMNETZ>.<2AF10E75-89F8-49A9-BD07-F50E394C401F@gmail.com>.5622.68.Xref: news.gmane.org gmane.emacs.orgmode:9209.
.
205 .

and then nothing more.

If I look at the packages, then the client side sends a FIN, ACK, and
then a RST, ACK, just before the connection terminates.

I haven't the TCP specification in my head, but this looks like the
client, ie. Gnus/Emacs, for some reason terminating the connection....? 

So why does it complain that the server closed the connection?  I guess
it could be the emacs TCP/IP support closing the connection, and Gnus
complaining that the server had closed the connection...?