Re: happy with 56k? & USR stuff

nezsez@iglobal.net
Fri, 29 Aug 1997 13:00:57 -0500 (CDT)

PM3's, comOS3.7, and the 56k modems have been very good running on
two PRI's from SWB here. The only problems we've had are with the
USR modems, and s15=128 (use MNP/MNP5 not 42/42.bis ) or &k0
( diable all datacompression )has fixed all those so far, and
some problems with the various versions of STAC in use ( we used
%c0 to disable some of the clients implementation )
We have a problem with one Zoom 56k modem, which ran fine for 7 days
(at 33.6 since it's pre 1.0) then all of the sudden won't negotiate.
Other than that, it seems the pri setup is about the same as the
pm3 with T-1 with the 33.6 digital modems. ComOS 3.7 generally
improved connect rates over 3.5b20, and seems to be at least as
stable. We also have not had some of the memory usage problems
others have reported, but we are not using OSPF or BGP on this
box either.

Livingston PortMaster PM-3 ComOS 3.7
System uptime is 9 days 21 hours 31 minutes

can anyone explain why using &k0 ( thus LAPM/NONE ) apparently
fixes the disconnect prob with USR's when it is supposed to be
an error correction problem with v.42bis? Using s15=128 to force
MNP/MNP5 by disabling v.42bis ( 42/42.bis) also solves the problem
but this makes sense. I know the EC/DC stuff is closely related/
intertwined in the code, but maybe the usr thing is really a
datacompression issue and not an error correction one.
IF this is so, then which would be better, LAPM with no datacompression
or mnp with mnp5 datacompression? The way I read it, LAPM is the newer
, better method, but datacompression is pretty much useless on todays
internet except for large text files, so it seems LAPM/NONE would
be the way to go.

On 29-Aug-97 Tim Hodges, Triangle Telephone Cooperative Assoc. wrote:

<snippage here>
>My hunch from reading the list is that PM-3's fed with
>PRI have MUCH better connect rates and better connections generally.
>
>Tim Hodges
>7thodges@3rivers.net
>
<snippage>
>
>At 12:13 PM 8/29/97 -0400, you wrote:
>>We downgraded all our units and are waiting for a comos revision before we
>>try implementing the 56k cards/3.7 comos again.
<snippage>
>>Chris Evoy