[1284] in linux-net channel archive
Re: SOLVED Re: Asymmetric TCP/PPP performance (very poor receiving TCP)
daemon@ATHENA.MIT.EDU (Al Longyear)
Sun Oct 29 02:20:09 1995
From: longyear@netcom.com (Al Longyear)
To: rlk@tiac.net (Robert L Krawitz)
Date: Sat, 28 Oct 1995 20:43:35 -0700 (PDT)
Cc: linux-net@vger.rutgers.edu
In-Reply-To: <199510272345.TAA24333@sunspot.tiac.net> from "Robert L Krawitz" at Oct 27, 95 07:45:51 pm
Robert L Krawitz wrote:
> It turns out that the Cisco 2509 at the other end of the PPP
> connection didn't like me trying to negotiate the MRU with it, and
> rejected my LCP configuration request (which included pcomp and
> accomp). Hence I needed -mru in my pppd command line. At this point
> I'm sending at 12K/sec and receiving at 10K. The remaining
> discrepancy needs further analysis.
(Check the asynmaps and ensure that your modems are really running at
the same speed. The V.34 modems may send at a speed which is different
than what they receive. Both speeds are adjusted independantly to
reflect the line quality.)
> Perhaps this should be added to the relevant HOWTO. The Cisco 2500 is
> a fairly new piece of hardware, but I suspect that it will be a fairly
> popular one since it's a good, cheap, flexible router.
Is that "2500" or "2509" or is the 2509 a member of the 2500 family
and would you know if other members have the same problem?
This is strange. There are a couple of 'reject' conditions within PPP.
The frame may have been rejected due to a protocol violation. This is
a LCP frame of the type 'protocol reject'. It ususally means that "I
don't know what this frame type is". The Linux PPP sends this to
Morningstar when they want LQR protocol since the code before 2.3alpha
does not understand LRQ.
You should not have seen this condition. It would mean that there is
something terribly wrong with Cisco's implementation of PPP since the
LCP frame is about the only 'manditory' protocol for PPP.
There is a second type of reject. This occurs in the LCP protcol
itself as the two peers are negotiating the options. This is the type
of 'reject' which you should have seen. It means something similar to
the protocol reject in that "I don't understand this option. Don't ask
it again."
If you did see a reject for the CI (configuration identifier) and the
Linux PPP code did not properly recover by doing the same thing as
"-mru" then please send me the debug trace. I will need to fix the
broken Linux PPP code.
If you received a protocol reject then someone needs to complain to
Cisco as their implementors will need to clean their glasses and
re-read the RFC documents.
Granted, I haven't had many implementations which wanted to reject the
MRU setting; but the Linux PPP code should have handled the condtion.
--
Al Longyear longyear@netcom.com
Finger for PGP key
