Path: utzoo!news-server.csri.toronto.edu!rutgers!usc!zaphod.mps.ohio-state.edu!caen!hellgate.utah.edu!dog.ee.lbl.gov!nosc!crash!simpact!jeh
From: jeh@dcs.simpact.com
Newsgroups: comp.dcom.modems
Subject: Re: Hayes Smartmodem 9600 troubles
Message-ID: <1991Mar8.090654.2210@dcs.simpact.com>
Date: 8 Mar 91 17:06:54 GMT
References: <1991Mar4.025731.21550@ping.chi.il.us> <3827.27d24b3a@hayes.uucp> <1991Mar5.093318.2146@llustig.palo-alto.ca.us>
Organization: Simpact Associates, San Diego CA
Lines: 33

In article <1991Mar5.093318.2146@llustig.palo-alto.ca.us>, david@llustig.palo-alto.ca.us (David Schachter) writes:
> In article <3827.27d24b3a@hayes.uucp> tnixon@hayes.uucp writes:
>>Like any other error-control modem, the VSM9600 performs best with a 
>>protocol that doesn't stop and wait for acknowledgements after every 
                                                                 ^^^^^
>>frame.  I'm not familiar enough with UUCP to suggest what settings 
>>or alternatives you might have to consider to make it work better.  
>>The VSM9600 generally works better than PEP with stop-and-wait 
>>protocols because its turnaround time is so much faster, but there 
>>will nevertheless be some degradation from optimum.
> 
> Toby is being a bit disingenuous here.  His "The VSM9600 generally works better
> than PEP with stop-and-wait protocols" is true for non-spoofed protocols.  The
> original interrogative was why the Hayes modem runs like a dead dog on a hot
> August day with uucp; for UUCP, PEP beats the Hayes by a country mile because
> PEP spoofs UUCP, converting into a non-stop-and-wait protocol.

Wait a second.  Uucp is NOT a stop-and-wait protocol by this definition (note
the highlighted word "every" above).  A stop-and-wait protocol has a transmit
window size of one; standard uucps have a transmit window size of three
64-character packets.  If you get the acks back quickly enough, uucp will
transmit continuously (within each file, anyway). 

Now, uucp at 3x64 effectively *becomes* "stop and wait" on high-speed links
with long end-to-end delays (even in the absence of turnaround delays), but
many uucps can be set to use a window of up to seven 4K packets, which should
give enough time for an ACK to come back for the first packet before the
seventh is sent even on VERY high speed, long-delay links. 

	--- Jamie Hanrahan, Simpact Associates, San Diego CA
uucp protocol guru, VMSnet [DECUS uucp] Working Group, US DECUS VAX Systems SIG
Internet:  jeh@dcs.simpact.com, or if that fails, jeh@crash.cts.com
Uucp:  ...{crash,scubed,decwrl}!simpact!jeh