Back around 1980, there was a low cost (by the standards of the time, or about $500 MSRP) networking system known as Corvus Omninet. Eventually, the price came down to a dealer cost of about $150 per card. This was a poor man's ethernet. I think cards were availible for Apple II, IBM PC, Mac, and S100 computers and maybe others. Each card had its own 6800 Family CPU, HDLC controller, firmware for the coprocessor, firmware for the host processoor, host interface (dual port memory), and the line driver/receiver. The physical layer was 1 megabit per second RS-485 HDLC syncronous serial.
Logically, the cards used the notion of a socket to communcate which was a very primitive cousin to berkeley (TCP/IP) sockets. In the corvus case, there were only a few possible sockets and the socket number corresponded to the address in shared memory where the data sent/received was buffered.
Historically, omninet came after Ethernet; however, ethernet was too expensive for widespread use with microcomputers and was primarily used with mainframes systems at the time. I think it may have come before Arcnet. It came before PC NET, Token Ring, and various others.
I managed a network of Apple II computers with Z80 cards running CP/M networked via Omninet with a Corvus Disk Server. Later IBM PCs were added to the network. Even later, we networked the PCs using Novel Netware over Omninet with a PC serving as a File Server.
Corvus products often had clever ideas rendered almost completely unusuable by corner cutting. If they could cut the cost by 10% and functinality by 90%, you can bet they would do it. Almost all Corvus products had fatal flaws. Clever network administrators would figure out ways around the problems.
This was the name of the network interface card. Apparently, it was designed by NEC.
This was the name of the corvus software. The login screen had a constellation of stars rendered using the asterisk character.
This is a Disk Server, not a File Server. Access was by sectors rather than by files. Any interpretation as files was created by the operating system. As a result, it was not possible to share files read/write. Typically, we had a shared read-only partition which contained DOS and various other software and a tiny (floppy sized) Read/write partition for each user which contained their files. It was about 20MB in size.
Corvus had originally sold external disk drives with their own proprietary interface and adapter cards allowing these drives to be connected to various different computers. Then they added the omninet disk server interface onto one of these drives. There is a copy of an advertisement at this page: http://www.spies.com/~aek/orphanage.html
The only way for one user to safely share files with another user via the disk server was via "pipes". These were FIFO queues that allowed one user to send a file to another. These were sort of like a primative form of email. Various glitches would cause corrupted pipes (whose space could not be reclaimed without erasing all pipes) so I wrote some utilities to manipulate pipes.
The Concept was a diskless workstation with a 68000 CPU 256/512KB ram, a rotatable display (portrait/landsccape). 4 apple II compatible expansion slots, flopy and winchester optional. 460x720 B/W graphics, 120x66 character text display. Released April 1982. For more info, including copies of manuals and boot prom listing, see : http://www.spies.com/~aek/orphanage.html
Corvus had a product that allowed you to backup your computer (or the Disk Server) to videotape using an ordiary VCR. No random access, of course. And they really screwed up the error correction. They simply repeated the block of data 4 times, consequetively. Unfortunately, the nature of video tape is such that if there is an error on one frame it is likely to affect succeeding frames. If they had done something like 1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4 instead of 1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4 your chances of recovering the data would have been much higher. Patent 4380047 (04/12/1983).
This was a continuous loop, 100 track, tape backup device. I think the capacity was around 200MB.
This was the tap box where the 15' cable from each workstation NIC connected to the main BUS. Guaranteed to fail due to poor quality insulation displacement connections and poor quality 1/8" phone jack. We replaced the phone jacks and soldered the connections.
Yeah, right. Security is entirely enforced by the client software instead of at the server level which means there is absolutely no security. The client software will only let the designated recipient read a pipe but with my utilities anyone can. The client software will only let authorized persons mount a disk partition but anyone can read the entire partition sector by sector using my utilities below.
The passwords are stored on the disk with a very crude symetric encryption algorithm. the usernames are also encrypted so a simple search of the disk for known usernames won't find the password repository but changing a username or password between two scans of the disk with CS.C will reveal the location. I don't have my notes on what the encryption algorithm is but I apparently had broken it by Jul 1984.
I wrote a number of utilities around 1984 for use with the old corvus disk fservers. Download Corvus Interface Disk 2 Corvus Interface Disk 2 (source and executables). These were compiled using the DeSmet C compiler. I think I wrote all of the C code here although there may be a few files from corvus (there are a few BASIC programs which they wrote). Files:
I reverse engineered (disassembled) the BIOS on the omninet cards. If I ever find the code, I will post it here.
Here are some messages I sent people which had information that may be of use to others.
From whitis@freelabs.com Thu Feb 21 18:58:30 2002
Date: Tue, 13 Nov 2001 21:33:32 -0500 (EST)
From: Mark Whitis <whitis@freelabs.com>
Subject: Re: Corvus
On Tue, 13 Nov 2001, Curt Vendel wrote:
> Hi,
>
> I ran across your resume while looking for old Corvus information.
> I own a Corvus Concept and some Corvus Transporter cards for the IBM
> PC and I'm searching for any technical data and software connected
> with these systems. If you have any old tech docs and software that
> you'd be willing to sell, please let me know, thanks.
Many years ago, I dissassembled the firmware on the corvus cards
and wrote some documentation on how to program them and wrote some
utility programs in C to do some stuff the corvus software didn't do.
It would take a lot of time to find those files if I still have them.
--
Mark Whitis http://www.freelabs.com/~whitis/ NO SPAM
Author of many open source software packages.
Coauthor: Linux Programming Unleashed (1st Edition)
From whitis@freelabs.com Thu Feb 21 18:58:30 2002
Date: Wed, 14 Nov 2001 17:51:36 -0500 (EST)
From: Mark Whitis <whitis@freelabs.com>
Subject: Re: Corvus
On Tue, 13 Nov 2001, Curt Vendel wrote:
> Mark,
>
> Thanks for the reply.... if you ever come across them and would be
> interested in parting them them, please keep me in mind, thanks much.
I think I found some disks with corvus stuff on them but I don't know
what is on them or whether it would be sufficient to use them.
Unfortunately, they are 5-1/2 inch disks so I can't read them
right now.
Corvus Omninet:
- 1 MB/s RS-485 multidrop bus network of up to 63 nodes
- HDLC based
- unsheilded twisted pair wire, differential signalling, not isolated
from ground.
- Disk server, not file server based (though novell
supported the omninet cards).
- As far as I know, no one ever wrote an omninet driver for a decent
networking suite (i.e. TCP/IP).
- Do not use their insulation piercing tap boxes without
soldering the connections. The connections are very crappy.
- Termination is required at each end of the bus (150 ohm?) per RS-485).
- Maximum wire length is something like 500-1000 feet (see RS-485 spec).
At least it was enough for us to wire 6 suites in two buildings
on one bus.
- each card supports multiple sockets (connection endpoints). socket
numbers are 16 bit integers like TCP/IP but only 4 are actually
supported with numbers (like B800) that suggest they might be
the address of a buffer in the omninet card's internal cpu's
address.
- Cards were availible for Apple II (typically used with a Z80
CPU card as well), S100?, and later IBM PC.
- The card had a cheep pin connector on it and a cable that
connected to that and had a 1/8" head phone plug which
connected to the tap boxes.
- The network was prone to lightning damage with long network
wires and pathetic surge protection on the cards. Zener diodes
from each half of the pair to ground help.
- when the RS-485 transceivers get blown up, you get strange
transmission reflection line problems. Machines A and B which
are sitting next to each other may be able to both talk to
machine C halfway across the building but can't talk to each
other.
I have no information on the corvus concept. I think that was a
68000 based computer.
Any software I might find will be DOS based, not windows let alone
a real operating system.
From your domain name, though, I get the impression you like tinkering
with antique computers rather than looking for a cheap networking
solution.
Incidently, I have three atari 1020 printer/plotters I was thinking
of selling on ebay.
> ----- Original Message -----
> From: "Mark Whitis" <whitis@freelabs.com>
> Sent: Tuesday, November 13, 2001 9:33 PM
> Subject: Re: Corvus
>
>
> >
> > > Hi,
> > >
> > > I ran across your resume while looking for old Corvus information.
> > > I own a Corvus Concept and some Corvus Transporter cards for the IBM
> > > PC and I'm searching for any technical data and software connected
> > > with these systems. If you have any old tech docs and software that
> > > you'd be willing to sell, please let me know, thanks.
> >
> > Many years ago, I dissassembled the firmware on the corvus cards
> > and wrote some documentation on how to program them and wrote some
> > utility programs in C to do some stuff the corvus software didn't do.
> >
> > It would take a lot of time to find those files if I still have them.
> >
> > --
> > Mark Whitis http://www.freelabs.com/~whitis/ NO SPAM
> > Author of many open source software packages.
> > Coauthor: Linux Programming Unleashed (1st Edition)
> >
>
--
--
Mark Whitis http://www.freelabs.com/~whitis/ NO SPAM
Author of many open source software packages.
Coauthor: Linux Programming Unleashed (1st Edition)
From whitis@freelabs.com Thu Feb 21 18:58:31 2002
Date: Sat, 19 Jan 2002 16:33:58 -0500 (EST)
From: Mark Whitis <whitis@freelabs.com>
Subject: Re: Corvus Systems Omninet 4 LAN
> Dear Mr. Whitis,
>
> [...]I am E-Mailing you
> because I have a problem I hope you can help me solve. I saw your
> resume on the internet while I was looking for information on the
> Corvus Systems Omninet 4 LAN. I saw that you have a great deal of
> experience with computer system including the Omninet.
>
> My problem is this... I have recently aquired an automatic scoring
> system manufactured by Dacos Worldwide. It uses the Omninet 4 LAN.
> I did not get a network cable with the system. I don't know the cable
> specifications I should use to build a new cable. Since these
> companies are gone, I have no support to call. The manual I have does
> not specify a cable but it does say to terminate the open ends with
> 150 ohm resistors.
>
> While both my brother-in-law and I have spent considerable time
> searching the internet, we have come up empty as far as a cable spec
> is conserned.
The 150 ohm termination suggests it is compatible with the version
of omninet I used.
The cable you need is:
- Configuration:
Resistor -----Tap----Tap----Tap----Tap---Tap---------- Resistor
| | | | |
| | | | |
| | | | |
Card Card Card Card Card
- adapter cables and bus cables were unsheilded twisted single pair with
150 ohm impedance. I think the bulk cable we used was manufactured
by belden. If you use sheilded cable, the capacitance will interfere
with proper operation. Running the wire through grounded metal
conduit, however, is probably ok and even a good thing.
- I think you will find connectors which mate with the connectors
on the card in the digikey catalog. They are pretty common.
- The original tap boxes had insulation displacement (wire piercing)
connections to the main bus cable. These ALWAYS failed and had
to be soldered.
- The tap boxes had mono 1/8" miniature phone jacks as the
connector for the cable to the adapter card. You can get
these (male and female) from digikey, radio shack, and other
electronics parts suppliers.
- Tap box:
any small plastic enclosure (from digikey or radio shack) will
do
- I don't remember if there were two or three pins on the
corvus adapter card. If there were three, only two were
used. One may have been a key. If it isn't obvious
which pin isn't used, you can look at the traces on the
card. The unused pin will either be disconnected
or connected to ground.
- I don't remember what the polarity was (which of the two
pins on the card connected to tip and which to ring) but
it isn't important as long as you are consistant and
don't mix your tap boxes with corvuses.
- These cards are very susceptible to lightning damage
if you have a long cable. Symptoms:
- Strange packet losses.
Two machines near each other can each talk to some other machine
far away but neither adjacent machine can talk to each other.
(weird transmission line reflection problems).
- Network card locks up if it isn't used for some period of time.
This is the Pathological AutoUNsave feature. If you have been
working on your document for more than an hour without saving
the connections, the network card will fail so you lose your
changes. Cycling power restores normal operation as long
as you use the card frequently.
If you are running a long cable inside a building or, worse, running
a cable between buildings or outside (such as to a stadium scoreboard)
you will need lightning surge protection.
For lightning protection, the technique I developed is to use two
5.1 volt zener diode to ground. For outdoor cable runs, more
elaborate protection is needed.
----------+------------------------ diff+
|
-------------------+---------------- diff-
| |
--- ---
/ \ / \ (zener)
--- ---
| |
| |
----- -----
--- --- (ground)
- -
- The signal is differential. Neither of the two wires connects to
ground. If you ground either side, you will have trouble.
> Do you know what cable I should use? I hope you can help!
>
> The model number of the Omninet cards is: 8010-13419-01 Rev B.
--
--
Mark Whitis http://www.freelabs.com/~whitis/ NO SPAM
Author of many open source software packages.
Coauthor: Linux Programming Unleashed (1st Edition)