Subject: Re: Localtalk interface design ideas
To: Bill Studenmund <wrstuden@loki.stanford.edu>
From: Mike O'Dell <mo@UU.NET>
List: tech-net
Date: 02/13/1997 20:55:40
wow, this is a blast from the past.

The Localtalk implementations have always been on Zilog 8530 USART
parts.  the packet format has an RTS/CTS transaction at the front,
and if that gets approved, then the packet streams.  that way the
collissions take place on the RTS/CTS exchange and kill tiny packets.
essentially a light-weight reservation protocol. I remember being
very impressed by Yogen's design when I first read the hand-xeroxed
pages of "Inside Appletalk" many, many moons ago....

and remember that the original Mac implementations were running
vvvvveeeeerrrrrryyyyy slow processors by today's standards.

on today's speedy processors, I don't see why a suitably tuned
ASM pseudo-DMA thing would be able to do byte-time interrupts
with no problems.  it is possible to get an 8530 to do DMA, but
it  isn't pretty, especially on an ISA bus.

	-mo

PS - C code to drive an 8530 in Localtalk is probably still FTPable
from somewhere at Stanford as part of Bill Croft's original
Appletalk/Ethernet gateway. the design was purchased and
commercialized by I think what became Shiva ("FastPath").
I remember looking around a year or two ago and saw the code
was still around. it was for a 68K, but stroking the 8530 is
pretty processor-neutral.