Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!mintaka!bloom-beacon!eru!hagbard!sunic!sics.se!fuug!demos!avg
From: avg@hq.demos.su (Vadim Antonov)
Newsgroups: comp.dcom.modems
Subject: Re: Telebit and V.32bis
Message-ID: <1991Jun10.175931.11466@hq.demos.su>
Date: 10 Jun 91 17:59:31 GMT
References: <98879@lll-winken.LLNL.GOV> <1991Jun6.193254.3886@robobar.co.uk> <1991Jun07.020706.16088@nstar.rn.com> <1991Jun7.193033.7604@robobar.co.uk>
Organization: DEMOS, Moscow, USSR
Lines: 44

In <1991Jun7.193033.7604@robobar.co.uk> ronald@robobar.co.uk (Ronald S H Khoo) writes:
>"PEP at 200 cps
>is better than no connection at all" is an important consideration.

>I really would like to know how well V.32is and its implementations
>would compare in situations like that.

Hm. Our team has sufficient experience fighting against poor quality
lines. PEP simply works. V.32 sucks. Cannot say much about V.32bis (never
tried) but my guess the result will be exactly the same. The problem
with adverse lines is that they may vary in a zillion of ways. Some
lines works fine with higher frequencies but have terrible 50Hz noise.
Some lines cannot pass higher tones. Some lines have awful echoes from
non-balanced intertrunk connections. Some comes through various multiplexing
devices. V.XX protocols require line able to carry well-defined band
of frequencies and practically cannot adopt their energy/frequency
distribution to the actual line conditions. Multiband nature of PEP
allows it.

Lowering speed of V.32bis connection cannot help to deal with line
noised at the carrier frequency (for example). PEP simply turnes this
band off. I can imagine a lot of realistic line conditions making V.32bis
useless while PEP still can push data. Actually I've seen a line V.22
sucks at but PEP works!

>If PEP wins, that means sticking to UUCP.  If V.32{,bis}
>wins, we get to switch to IP.  None the less, if PEP wins, I'll continue
>to buy PEP, because reliable UUCP is much better than unreliable IP.

Wrong. Of course, PEP shows huge turnaround time making interactive
sessions painful (but not making it impossible at all). Surprisingly
file transfer over PEP/CSLIP is not much worse than over PEP/UUCP (about
15% loss).

What I really wish to see is true duplex PEP - sharing the part of
channels between directions. Dynamically, of course. I think the only
reason why PEP is half-duplex is simple - transiever (sp?) scheme
requires only one DSP switched from transmit to receiving. Duplex
scheme would take a couple of DSPs or one with more units inside.
Still I hope Telebit will make such a beast - to defend these horses
designed by a committee.

Vadim Antonov
DEMOS, Moscow, USSR