George Fliger wrote in a message to Mike Bilow:
 MB> My PC networking days go back to truly ancient kludges such as 
 MB> the Corvus Omninet -- remember those?
 GF> My God!  Somebody actually remembers those???  I thought I
 GF> was the only one who dealt with that misfortune. 
Yes, and not fondly.  I first had to deal with the Corvus boxes in 1980 or 
so, when they were introduced as shared hard drives for TRS-80 networks.  
Corvus supplied two big white boxes, one containing a 10 MB hard drive and 
power supply, and the other containing a 16-port multiplexer for the TRS-80 
expansion interface bus.  It was a horrible mess, but it worked for years.  
The Corvus Ominet was actually a later variation for the IBM PC.
The fundamental idea behind the Corvus architecture, that of electrically 
extending machine buses with differential drivers and ribbon cables, has 
mercifully been totally forgotten.  As soon as it became economical to use 
fairly intelligent microcontrollers on cards, the inherent advantages of real 
networking systems such as token ring and Ethernet became obvious.
I was a registered developer (or whatever it was called) for the TRS-80, 
which Tandy was then actively promoting as an ideal business solution.  That 
was not a very good idea, since it took them a while to make such 
conveniences as printer ports and serial ports standard equipment, and they 
never did implement a competent serial port.  In fact, we finally designed 
our own serial port boards for the TRS-80, and I wrote what would pass for 
drivers.  We used the 8250, then a $50 chip that was hard to find in anything 
other than the Heath/Zenith line of Z-80 machines, but our experience with it 
came in very handy when IBM later chose it for their PC, which also drove the 
price down to about $15.
 GF> You probably pre-date me networking wise (no shot at your age, 
 GF> really!).
I started dealing with microcomputers around 1977.  In fact, I was at the 
show where the TRS-80 was introduced, and it was the talk of the place 
because it cost only $600 and you did not have to solder it yourself.  It was 
also the first real coup by Microsoft, since their BASIC was in the system 
ROM.  Until the TRS-80, Microsoft was just one among several players in the 
BASIC market, and they were probably not the best technically.
One serious contender for technical supremacy was Benton Harbor BASIC, 
written by Gordon Letwin to run under his HDOS operating system on the 
Heath/Zenith machines, but Heath would not license it to competitors.  BH 
BASIC even had the beginning of dynamic memory management, a real innovation 
in a 64 KB address space.  Gordon Letwin later became frustrated with Heath 
and joined Microsoft as employee number 10 or so, eventually inventing OS/2.
 GF> I can remember the first Netware system I
 GF> installed back in 1986.  IBM PC XT's and AT's (35
 GF> workstations total) with 2.5 Megs expanded memory, Token
 GF> Ring NICs and QuarterDeck's DesqView for a front-end.  The
 GF> server was an AT (6Mhz) with two Seagate 4096 80Meg MFM
 GF> drives.  The whole bundle (including the programmer time)
 GF> cost the company around $250,000 to imple- ment.  And to
 GF> think, all they had for the entire company was a bit less
 GF> than 160Megs total server space.  They never thought they'd
 GF> run out. :) 
NetWare evolved very strangely.  I think someone here mentioned a couple of 
days ago the original system which depended upon a dedicated 68000 CPU box.  
By 1986, though, things had greatly matured and NetWare was widely used.  The 
Seagate ST-4096 was among the larger drives you could get, but had a tendency 
to develop mispositioning error, so we stayed away from it.  I do still have 
three or four of them sitting around, though.  We moved to Maxtor, which in 
those days made some of the finest and most reliable stuff around, although I 
certainly would not say the same about them today.  For many years, my 
personal desktop ran a Maxtor 180 MB drive that we pulled from a file server 
because it "sounded funny," but we couldn't send it back for that -- and it 
did run flawlessly for about five years, 24-hours a day, in my desktop 
machine until it was retired in favor of a larger drive.
The largest network we did in that era had three NetWare servers and about 
300 workstations, and it was also token ring since Ethernet was not really a 
proven technology for more than a few workstations, nor was coax practical 
for many physical office layouts.  We did use Ethernet for the interserver 
backbone, which even then we had the good sense to keep separate.  Each 
server had about 400 MB using Seagate ST-4383E ESDI drives, usually running 
on Western Digital WD1007 ESDI controllers.  It took about 4 MB RAM to cache 
these drives, and both the drive capacity and the physical memory were 
considered enormous for the time.
 
-- Mike
--- 
---------------