Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!ncar!ames!pasteur!ucbvax!unisoft!paul
From: paul@unisoft.UUCP (n)
Newsgroups: comp.protocols.appletalk
Subject: Re: Speed of LocalTalk, gateways
Message-ID: <1347@unisoft.UUCP>
Date: 25 Oct 88 17:04:03 GMT
References: <964*kenw@noah.arc.cdn> <639@kinetics.UUCP>
Reply-To: paul@unisoft.UUCP (Paul Campbell)
Lines: 37

In article <639@kinetics.UUCP> minshall@kinetics.UUCP (Greg Minshall) writes:
>From article <964*kenw@noah.arc.cdn>, by kenw@noah.arc.CDN (Ken Wallewein):
>> 
>>   I keep hearing about people saying LocalTalk is slow.  Frankly, I'm 
>> skeptical.  
>
>	I've seen LocalTalk (thru a FastPath) run up to 18 Kbytes/second
>(of "useful" data).  Needless to say, there is more to performance than just
>the speed of the wire.  For example, block size, window size, etc., all
>have a large effect.

I run UUCP over LocalTalk (don't ask why :-) .... it runs on top of ATP,
it's a relatively noise-free net so retries are at a minimum. I sometimes see
data rates over 100k. The main reason for this is that I can do 8
(large) packet writes across the net (I also have hardware that will buffer
upto 25 Localtalk packets so I can do disk IO at the same time as transmission
and reception, and the Unix buffer cache etc that does delayed writes and
read ahead for me).

The main reason for posting this is to point out that it is possible to 
claim a large part of the LocalTalk bandwidth without really trying hard (in
fact I spent some time thinking about ways to detect when the net was getting
full so UUCP could throttle itself down a bit - it doesn't really care about
real-time response - anyone got any ideas - I thought of collecting collision
rate information and averaging over time).

	Paul

PS: if anyone else is crazy enough to want to run UUCP over AppleTalk, I'm
	willing to give away my sources (at least the parts that don't belong
	to AT&T - ie the ones for the protocol) just so we don't end up with
	multiple protocols for doing the same thing
-- 
Paul Campbell, UniSoft Corp. 6121 Hollis, Emeryville, Ca ..ucbvax!unisoft!paul  
Nothing here represents the opinions of UniSoft or its employees (except me)

	"Where was George?" - Nudge, nudge say no more