Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!lll-winken!gauss.llnl.gov!casey
From: casey@gauss.llnl.gov (Casey Leedom)
Newsgroups: comp.dcom.modems
Subject: Re: Overseas Connections with high-speed modems
Message-ID: <92421@lll-winken.LLNL.GOV>
Date: 4 Mar 91 00:15:41 GMT
References: <1018@nih-csl.nih.gov> <242@rwing.UUCP>
Sender: usenet@lll-winken.LLNL.GOV
Organization: Lawrence Livermore National Laboratory
Lines: 55
Nntp-Posting-Host: gauss.llnl.gov

/ From: bert@helix.nih.gov (Bert Tyler)
| 
| Apparently I did not emphasize enough that these folks will be using
| MS-DOS machines and MS-DOS-based software ... I received several
| responses about the advantages of Telebit's "spoofing" capability.  ...
\ Telebit modems won't find any protocol to "spoof".

/ From: pat@rwing.UUCP (Pat Myrto)
| 
| Oh, yes there ARE protocols to "spoof" in the MSDOS world - virtually all
| the non-streaming (ACK-NAK, or 'send and wait') file transfer protocols
| take a big performance hit with the hi-speed modems, even with non-PEP
| modem protocols, probably because of the packet-type nature of the
\ error-correction schemes, such as MNP.

  MNP and V.42 packetization is only one source of latency in the overall
communications link.  There's the time necessary for the application to
turn an ACK around, time to get the bits in and out of serial ports,
possibly even terminal servers and Ethernet links, etc.

  Fundamentally the bug is in the non-streaming protocol.  Anyone who
develops a protocol now that doesn't support sliding windows should be
removed from the gene pool.  It's just stupid not to support sliding
windows when the algorithms are now so well known.

  Unfortunately there are a lot of older protocols and, yes, stupid newer
protocols, that some people have to use.  It's unreasonable to expect
Telebit or any modem company to support spoofing of every such protocol
in the universe.  They can only afford to go after the more popular
ones.

  Therefore, if you have to use one of those stupid protocols and no
modem is available to spoof it, you'd better look around for the
combination of equipment and lower layer protocols that minimize your
latency in order to boost your over all performance.

  From a practical standpoint, I've noticed that my Telebit T2500 seems
to over a lower latency path when using V.42/V.42bis than with MNP5 --
both over a V.32 carrier.  I suspect that this is probably because I've
heard that the default packet size for MNP5 is 256 bytes and the default
packet size for V.42 is 128 bytes.  This reduces the streaming throughput
rates for V.42 but also decreases the round trip latency for
non-streaming protocols like XMODEM.  Some modems may allow you to set
the packet size in MNP or V.42.  The T2500, as far as I can see, does not
have any way of doing this.  (By the way, I can hardly wait till we see
modems whose link layers dynamically negotiate their packet sizes based
on a combination of Nagle algorithm methods and analysis of burst error
rates (BERTS) on the link.)

  Also from a practical standpoint, Telebit's PEP protocol is going to
give you very large latency.  This will be especially true if the
protocols your are using have large ACK packets or allow data
transmission in both directions at once.

Casey