The key to an easily networkable personal computer in 1983 was whether or not it had a well-documented expansion bus. An expansion bus is a socket or series of sockets on the main computer circuit board (the mother board) to which other circuit boards (daughter boards) can be attached to expand the computer’s capabilities. This is the idea of modular design applied to personal computers.
The first generation personal computers were built for hobbyists, with an expansion bus on the mother board. The uses to which these early computers would be put were so varied that the designers needed to make them easy to customize. Game playing, voice recognition, appliance control, and serious number crunching were a few of the many uses the hobbyist users of the early ’80s found for personal computers. The easy way to customize was to add different boards, so the early makers always put in a place where many different kinds of boards could be added—the expansion bus.
An expansion bus can be a single socket—as it was on the Macintosh SE—or it can be many. The most prolific were the late ’70s era IMSAIs and Cromemcos which had 19 and 21 slots respectively. These were so modular, in fact, that the mother board consisted of nothing but expansion slots—all the RAM, all the I/O, and even the CPU came on daughter cards.
But modularity adds expense and size to products. By 1982 some of the next generation CP/M-based computers that were coming out were integrated—they had everything on the mother board and no expansion bus. These new machines cost less and looked more stylish, but they couldn’t accept a networking card.
The IBM PC followed the Apple II tradition and stayed modular. This made the PC exciting to people like Drew who wanted to do more with a computer than just connect it to a printer and a modem.
Novell’s S-Net hardware was a product of its times. Engineers and marketers had discussed other alternatives but S-Net seemed best suited.
In 1981–82 there were no networking standards. The computer magazine articles of the day were still discussing the theoretical virtues of baseband versus broadband (modulated versus unmodulated) signal transmission, and various physical networking topologies such as linear bus and star, because there were few real networks to discuss.
Novell considered Ethernet, but in those days it was much too expensive—over $1,000 per connection—and “thin” Ethernet hadn’t been developed, so the only way to connect was with bulky converter boxes connecting equally bulky quarter-inch-thick coaxial cable.
ARCnet was also explored and rejected because of its cost.
The end result of this elimination process was S-Net, an in-house design. S-Net used a star topology with all the workstations connecting directly to the file server. It used an RS-422 electrical protocol that was inexpensive to design and build, because it was carried over dual twisted pair cable (alias 4-wire phone cable).
S-Net was attractive for its simplicity. All the cables had only two devices on them. This meant there was no need for a sophisticated communications protocol[Footnote 1] that had to provide for addressing different stations on the wire. The file server knew which workstation it was talking to simply by knowing which communications port the signal came from.
S-Net was yet another quick and inexpensive choice in the headlong rush to get the LAN to the market. It was a good choice at the time, when LANs were self-contained systems of half a dozen workstations devoted to sharing hard disks and printers, and little more. S‑Net would serve less well as LANs embraced larger networks and connectivity to many other kinds of data processing systems.
| Footnote 1: The Internet’s protocol suite (TCP/IP, for Transmission Control Protocol and Internet Protocol) is an example of the sort of sophisticated protocol that was avoided, along with the army of programmers that would have been necessary to produce it. |