LinuxSA Mailing list archives

  From: Alan Kennington <akenning@dog.topology.org>
  To  : linux@senet.com.au
  Date: Thu, 4 Mar 1999 21:01:21 +1030

Re: Modem problems fix - Re: pppd delay (fwd)

Benjamin,

Thanks for that note about the bsdcomp parameter in pppd.
At first I thought this was a ray of light.
But as Darryl said, the problem occurs with kernel 2.0.36.

There's another difficulty with the proposal.
pppd runs packets over a byte-channel.
If it was a software problem in pppd, then there would
not be any modem re-training.

The engineer at the Telstra end said he was observing the
modem at his end re-training for 7 out of every 8.5
seconds. This is at a layer below the pppd layer, which sits
on top of the byte-layer provided by the modems,
which as Mark Newton pointed out is just a byte transport
machine. 
pppd compression, as far as I know, just chooses which
bytes to send, and it does not affect the modem ADC/DAC functions.

The reason this thing is so difficult to diagnose is the
intermittent nature of the problem.
Last time it took a month to appear.
I made a change to the parameters, and now I must wait
for over a month to be reasonably sure that it has worked.

Actually, where I work we have the equipment to really diagnose this
problem -- we have all kinds of spectrum analysis equipment,
and better-than-average oscilloscopes etc.
But I would be stunned if they agreed to let me 
stick that onto my phone line for a month to see
what's going on.

The observation of the Telstra engineer that the modem at their
end was re-training is the one thing that
makes me think that we are talking about a modem problem here.
Therefore in a sense it is a non-linux problem.
It _is_ a linux problem, though, in the sense that
machines that keep a ppp line up on a phone line for
longer than 1 month are more likely to be linux than
and non-unix OS.

Another possibility perhaps, is that I should be excluding
some control characters with the pppd async parameter
because there might be some character combinations mucking up the
modem. The dreaded +++ sequence is not a likely cause, by the
way, because it needs an idle time on either side.

Cheers,
Alan Kennington.

-- 
Check out the LinuxSA web pages at http://www.linuxsa.org.au/
To unsubscribe from the LinuxSA list:
  mail linuxsa-request@linuxsa.org.au with "unsubscribe" as the subject