Re: bug? gnus unreliable for sympatico.ca news server

Re: bug? gnus unreliable for sympatico.ca news server

[kgreiner]

* Tong * <sun_tong_001@yahoo.com> writes:

> Hi,
>
> Having newly moved to sympatico.ca news server, I found that gnus is
> extremely unreliable working under it. Sometime it works great, but
> most of the times, it just hung there -- the entire emacs doesn't
> respond to anything, even x-refresh request from window manager. I'm
> using gnus that is shipped with RH8. Meanwhile, Mozilla that comes
> with RH8 does not have any problem with the sympatico news server.
>
> It took me the entire afternoon and evening to trace down what could
> be wrong. I installed the nntpcache daemon, and watched closely the
> request (and respond) from Mozilla (and gnus) to/from the sympatico
> news server.
>
> When Mozilla requests a news article, it issues the following 3 commands:
>
> <- MODE READER
> <- GROUP comp.emacs
> <- ARTICLE 103393
>
> When gnus requests a news article however, it issues only the
> 'ARTICLE' command. This is where the problem is. The reason is as
> follows.
>
> The sympatico guys, in the effort to stop people fooling around with
> their news server, set a very short timeout limit between 'ARTICLE'
> command and 'GROUP' command. This is OK for Mozilla, but not OK for
> gnus, since gnus only issues 'GROUP' command once the group is
> selected. I now think that the reason why sometimes I can get
> everything working is because I happened to make the requests back to
> back, with minimum time delay in between. But when gnus has rested for
> a while, it will fail to request any further articles. The conclusion
> might be wrong here. It is based on my (careful) observation. I didn't
> open up gnus source to verify.
>
> Another thing I found about gnus is that when it hangs, I can break it
> out with ^G. But the problem is that there might be some internal
> state that is not cleared up, because further article requests will
> not be responded (no messages logs in messages buffer).
>
> I don't know who else are using the sympatico.ca news server, and
> whether they are suffering the same glitches that I am
> experiencing. Hope my effort (the whole afternoon and evening) can
> help gnus to improve. And hrrr, Anybody know a quick fix?
>
> thanks
>
> tong
>
>
> PS. the sympatico news server:
>
> NNTP server news1.on.sympatico.ca:
> 200 Bell Nexxia News Server (Twister v1.2.0)
>
> PPS. I'm not sure, but if you want to duplicate the reslult,
> installing the nntpcache daemon and set a very short connection
> timeout (5s) may help.

Thanks for the info.

The latest version of Oort gnus should fix this problem.  As I added a
timer to break out of hangs by closing the current connection then
opening a new connection and re-issuing the current command.

I'll look into added the GROUP command right before every ARTICLE as
that would help avoid unnecessary connection teardown and rebuild.

Kevin

Re: bug? gnus unreliable for sympatico.ca news server

[Lars Magne Ingebrigtsen]

* Tong * <sun_tong_001@yahoo.com> writes:

> oh, forget to ask, this "latest version", where to get it, from CVS or
> it is already wrapped up for download? If yes, is it
> <URL:http://quimby.gnus.org/gnus/dist/ognus-0.10.tar.gz>?

It's now up to 0.12.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: bug? gnus unreliable for sympatico.ca news server

[Kevin Greiner]

* Tong * <sun_tong_001@yahoo.com> writes:

> Thanks for such a quick respond.
>
> Kevin Greiner wrote:
>>>The latest version of Oort gnus should fix this problem.  As I added a
>>>timer to break out of hangs by closing the current connection then
>>>opening a new connection and re-issuing the current command.
>
> latest version of Oort gnus has already a workaround for my problem?
> Great news.
>
>>>I'll look into added the GROUP command right before every ARTICLE as
>>>that would help avoid unnecessary connection teardown and rebuild.
>
> Great!
>
>> Sorry forgot the most important part.  The new code does a better job
>> of recovering from sockets closed by the NNTP server.  However, it can
>> hang if the server chooses to simply ignore a connection (I have one
>> server that closes idle connections while another ignores them on the
>> theory that they've hung a resource-hungry client).  To avoid the
>> hang, you need to customize nntp-connection-timeout.  The value is in
>> seconds.  I've used 30 to good effect for a couple of months now.
>> For what you are describing, you probably won't need
>> nntp-connection-timeout.  Still it's there if you need it.
>
> Does it 'nntp-connection-timeout' correspond to the timer you just
> mentioned, to break out of hangs?

Yes.

> Then I guess I need it, 'cause seems to me sympatico.ca news server
> just ignore my (postponded) ARTICLE request, and hang my telnet
> session. All other commands, like xover, list, etc, will get a respond
> very fast, even after idle for a long time... Hope new gnus version
> can handle situations like this. (A group command should be
> enough. :-> )

Let me how this works out for you.  I use three different servers and
all three have different behaviors when a connection is idled.  I'd
like to use a general purpose solution but sympatico may fall outside
of the solution as it is currently written.

Kevin