Computer Chronicles Revisited, Part 96 — AppleShare, EtherTalk, and the AppleTalk Personal Network

Apple had a few false starts when it came to local area networks in the 1980s. In February 1983, the company announced AppleNet, which it described as a “scaled-down version” of Bob Metcalfe’s Ethernet standard. The idea was that AppleNet would offer a decentralized LAN that didn’t require a full server and could connect up to 128 Apple II, Apple III, or Lisa systems across a distance of up to 2,000 feet. Apple claimed this approach would make its LAN more affordable, with a per-node connection cost of under $500. The trade-off was that AppleNet would be slower than other Ethernet LANs.

Apple promised that AppleNet would ship sometime in late 1983. But that never came to pass. The system was seen for the first–and apparently only time–in public at the June 1983 National Computer Conference in Los Angeles, California. But four months later, in October 1983, Apple said “never mind” and canceled AppleNet.

There were a couple of reasons for the about-face. First, AppleNet was tied to the Lisa, which was also introduced in 1983 and quickly proved to be an expensive failure. Second, there was still uncertainty within Apple as to which networking standard to use. Apple chairman and co-founder Steve Jobs believed that Ethernet was the wrong way to go and felt the company should instead back IBM’s Token Ring networking standard. (In fairness to Jobs, a lot of people thought IBM would win this battle.) So after the Macintosh launched in early 1984, Apple announced AppleBus, which made it possible to create a LAN of up to 32 Macs or Lisas over 1,000 feet using serial cables and a $25 add-on box. Jobs dubbed this low-cost solution “free net” to the press.

Jobs also promised that AppleBus would be fully compatible with Token Ring. Of course, he said this before Token Ring was actually on the market. As InfoWorld’s Jim Bartimo noted, “How does Apple plan to be compatible with a network that doesn’t exist yet?”

Indeed, Token Ring didn’t ship until late 1985. So Apple needed to revise its networking plans yet again. In January 1985, the company announced its “Macintosh Office” strategy. This was based on the existing AppleBus protocol, which had been renamed AppleTalk (and later AppleTalk Personal Network). Macintosh Office initially focused on networking multiple Macs to the LaserWriter printer. But there would also be a central file server equipped with a 20- or 40-MB hard disk drive. This was important because the original Macs did not ship with hard drives.

But the file server wasn’t actually ready in January 1985. Apple promised it would ship that fall. Except that didn’t happen either. As 1985 bled into 1986–and Apple CEO John Sculley prevailed in his internal power struggle over Jobs–there were yet more delays in actually shipping a file server. It was not until January 1987 that Sculley announced (or rather re-announced) that an AppleTalk file server would finally be ready that May.

The new server, dubbed AppleShare, actually did ship, although not as a standalone piece of hardware. Rather, AppleShare was software that could be installed on an existing Mac to convert it into a dedicated file server. In May 1987, Apple also announced EtherTalk, an Ethernet adapter that made it possible to link IBM PCs and compatibles to an AppleShare server.

Which brings us to this November 1987 Computer Chronicles episode looking at how Apple’s fancy new networking setup worked. This was the second in a two-part series, with the previous episode focusing on lower-cost LAN solutions for the PC market. The episode began with Stewart Cheifet and co-host George Morrow demonstrating one of the “original” forms of networking–talking over two cans connected by a string. On a more serious note, Cheifet said that a better analogy for a computer network was the telephone system. Morrow agreed, noting the original phone system was based a “party line” network where you had multiple telephones connected to the same line. That was similar to the bus structure of today’s LANs, where multiple computers were connected to the same wire. And much like when there were multiple telephone companies that made it difficult to call between different systems, there were now multiple competing standards for computer networks, making it difficult for different machines to talk to one another.

Big Eight Accounting Firm Beta Tested Hybrid Network

Wendy Woods presented her first remote segment, narrating B-roll footage taken at the San Jose, California, branch of the accounting firm Arthur Young & Company, then one of the so-called Big Eight financial services firms. Woods said the firm faced some “special problems” when it came to building an integrated computer network. Incompatible hardware had to be divided into incompatible networks on three different floors of the office. On the eighth floor, Macintoshes were everywhere, but on the lower floors, IBM compatibles had the upper hand.

Today, Woods said the company was tasting a new gateway that allowed any user to communicate with any part of the network. Chris Veal, Arthur Young’s western regional director, told Woods that they were addressing the ability to take an existing network for a single environment–DOS–and gateway that into the Macintosh world. That meant using a dedicated MS-DOS machine that had both ARCnet and an AppleTalk cards. (ARCnet was an older networking standard that predated the IBM PC and Ethernet.) This machine could therefore talk to both networks and effectively created a seamless network on either the PC or Macintosh side.

Woods said that 3Com file servers presently linked different machines and different networks over telephone lines while the so-called interbridge talked to all of the networks. But the decision to keep different machines led to frustrations and delays. Veal said the biggest problem was coming to terms with the true capabilities of the network and not pushing the system beyond what it was expected to do. The problem used to be things like corrupted data and file allocation tables going out of whack. Now, the main problem was expending a lot of time trying something that was never intended to work and dealing with the consequences if it didn’t work.

Indeed, Woods note that the Mac-to-IBM gateway was not yet perfected. It took several tries to transfer a file for an on-air demonstration. But the appeal of the concept was undeniable. Through the interbridge, a user could retrieve a formatted document or send a memo or electronic mail anywhere on the network.

Woods added that at Arthur Young, Macintoshes outnumbered IBMs, which was still an unusual sight in the business world. It was common to see accountants lugging Macs back into their offices from field trips to see clients. And despite some persistent bugs in the IBM-to-Mac network, the firm planned to expand its system to provide network access from remote locations–and ultimately a wide-area network to link offices around the world.

AppleShare Turned Mac II Into File Server

Chris Bryant, the workgroups manager at Apple, joined Cheifet and Morrow for the next two segments. This first segment focused on AppleShare. Morrow opened by noting that networking was built-in to the Mac when it first came out but a lot had happened since the January 1984 launch. Where did things stand now with the Macintosh II? Bryant said the company had moved from the initial implementation of the AppleTalk Personal Network to add a number of additional services, including the AppleShare file server.

Bryant said that one of the first things that users noticed about file servers was that they were different to use than their local computers. But one of Apple’s priorities was making the AppleShare server as transparent as possible to the user, i.e., to make it as close to using their local machine as possible.

Bryant then explained the setup for his demonstration of the AppleShare file server. On the desk in front of him was a Macintosh II workstation with a color monitor that was connected to the file server over the network. On the other side of the studio, another Apple executive operated a second Macintosh II running AppleShare, a LaserWriter Plus printer, and a Macintosh SE workstation that was also connected to the server.

On his Macintosh II, Bryant opened the file manager to show the files and subdirectories on his local hard disk. He clicked the icon for the AppleShare volume, which opened a second window displaying the contents of the AppleShare server. He noted that the representation of both the local and network drives were nearly identical.

Morrow pointed out that one of the folder icons in the AppleShare window was gray. Bryant said that was a metaphor for resources that the user could not access. Cheifet asked why the user didn’t have access. Bryant said it was because the “owner” who had control over giving access kept that file private. In fact, he pointed out this particular gray icon represented a “drop” folder. The user could make changes to that folder but not see what was inside. This was useful in a situation like submitting status reports to a manager. Employees could add their reports to the folder, but only the manager could see them.

Morrow observed that a user-friendly system would make it possible to move an item from the local window to the file server window. Was that possible? Bryant said it was and he did just that. Morrow was suitably impressed. Cheifet asked for a reverse demonstration from the operator on the second workstation. Bryant showed that when the operator created a file on his machine, the icon then showed up on Bryant’s workstation. The other operator also showed how you could control access to a file using a separate menu.

Morrow asked how all of this information appeared on the file server. Was it the reverse of how it looked on the local workstation? Bryant noted the file server was dedicated to that purpose. A dedicated server maximized network performance for users. The server could run additional services such as LaserShare–Apple’s print spooling software for the LaserWriter–but not user applications. Bryant added that the decision to implement the file server on standard Macintosh hardware was a plus because as users needs grew they could easily upgrade the server.

Cheifet asked if you had to use a Macintosh II as the file server. Bryant said that AppleShare would run on a Macintosh Plus or a Macintosh SE, but the Macintosh II offered better performance and expandibility. Morrow clarified there was no structural difference between the Macintosh II used as a file server and one used as a workstation. Could you use a single machine as both? Bryant said it was possible but there would be issues, especially when trying to run applications on a server.

EtherTalk Connected Macs and PCs

For the next segment, Bryant showed how you could connect a Macintosh to an IBM PC or compatible. Cheifet asked about the networking protocol. Bryant said the Macintosh II on the desk was connected to the file server using EtherTalk. On the other side of the studio, a Macintosh SE and a Compaq Portable II were connected using the AppleTalk Personal Network, which was a lower-cost implementation with lower speed. Bryant explained that the AppleTalk protocols were designed into every Macintosh and LaserWriter. So the user interface was consistent across the Macintosh line regardless of the protocol used.

Cheifet asked for a demonstration of the Mac and Compaq Portable II working together. As a second operator worked the Compaq, Bryant explained that when a PC user accessed the Apple file server, it saw the directory structure as if it were a DOS machine. (In other words, it represented the file system using letters for the disk drives.) Bryant noted that the system even automatically shortened the names of longer Macintosh files to comply with the more limited 8.3 MS-DOS nomenclature structure.

The operator launched Lotus 1-2-3 on the Compaq and made changes to a spreadsheet. Bryant said he could then open that same document in Microsoft Excel on the Macintosh II. Excel automatically converted the file. Bryant then created a graph in Excel using the converted data.

Morrow asked about the cost of this approach to networking. Bryant reiterated that the hardware and software protocols required for networking were built-in to every Macintosh and LaserWriter. All you needed to connect workstations over the AppleTalk Personal Network was a $75 connector per node.

Stanford Building the First Blocks of the Internet

Wendy Woods presented her second remote segment from Stanford University in Palo Alto, California. Woods noted that Stanford had one of the largest campus computer networks in the United States, comprised of some 1,300 computer ranging from Macintoshes, sophisticated workstations and PCs, to Digital Equipment Corporation VAXes, MicroVAXes, and IBM mainframes.

All of these machines talked to one another via the Ethernet-based network, Woods said. In various closets across campus you could see the gateways and repeaters–in lay terms, the “traffic cops” that made the system across the vast distances requires. This network even extended into the dormitories, where students could access bulletin boards, job banks, and even the full bibliographic holdings of the Stanford University library system. Students could also do assigned courses in programming and send e-mail to instructors.

Bill Yundt, Stanford’s director of networking, told Woods that tying all of the diverse systems together required some trade-offs. He said the university had to write software for the Macintosh that ran in a compatible fashion with the other machines on the network using the TCP/IP networking protocol. That allowed a Macintosh to sit on an AppleTalk network and connect to the Ethernet-based networks such that the Mac looked like it was physically connected.

Woods noted that Macintoshes were slow and only had single-tasking operating systems. But students loved them. They were easy to operate and good teaching machines. The most common use for Macs were in classrooms, where small local area networks could share software programs stored on a dedicated file server. So far, French, history, and computer literacy were taught this way.

Linking all computers on campuses was an increasing phenomenon throughout the nation’s top universities, Woods said. MIT and Carnegie Mellon were two other examples. But Stanford officials said they were ahead of the rest because they found a way to make their network useful for top researchers and freshman students alike.

Was the “Age of Intelligent Network Managers” Coming?

For the final segment, regular contributors Jan Lewis and Tim Bajarin joined Cheifet and Morrow. Cheifet noted that there were still a lot of user complaints about the complexity, speed, and reliability of local area networks. Had we actually made much progress in this area? Lewis said we had, although there was a still tremendous way to go. It also came down to perspective. If you worked in an environment with a lot of different architectures and machines, the types of connections we had now were still inadequate. On the other hand, if you were in a department where some people used IBMs and some used Macs, a product like AppleShare allowed them to share files in a way that was not possible a couple of years ago. Lewis added that Apple’s implementation of AppleShare was truly elegant in terms of the user interface.

Bajarin said AppleShare was an indication of where things were headed in the future in terms of developing “intelligent network managers.” In order to see this kind of inter-connectivity work, we would need to see LANs working closely together with a much cleaner user interface and a lot of power. He agreed with Lewis that perspective also mattered. If you came from a mainframe or minicomputer environment, than today’s LANs had not come far enough. But if you were in a microcomputer environment where you couldn’t connect different machines at all two years ago, these LANs looked wonderful.

Morrow asked if Microsoft’s forthcoming Presentation Manager for OS/2 would offer an environment that makes LANs for the PC as natural as what we’ve seen from Apple? Bajarin said probably, although the big issue there would be memory. Software like Presentation Manager and Windows required a tremendous amount of both memory and CPU power to function together with networking protocols.

Apple Opted for More Expensive “Authoritarian” Approach to Networking

As I discussed in the introduction, it was a long road getting from Apple’s 1983 promise of a low-cost, low-speed LAN to shipping AppleShare more than four years later. Not that John Sculley didn’t take a victory lap after crossing the finish line. Announcing the re-launch at Jonathan Seybold’s first Desktop Communications Conference in late January 1987, Sculley proclaimed that AppleShare was “just as significant as the original Macintosh” and that the project’s chief engineer, Gursharan Sidhu, would be “recognized as a folk hero for what he has done.”

AppleShare was definitely the talk of the Seybold Conference. But not all the talk was positive. John Markoff, writing for the San Francisco Examiner, said the “self-congratulatory mood” surrounding Apple “tended to obscure” the substantial delays in the project. Markoff said Bill Gates, a “staunch Apple ally,” noted that “while Apple has a clear lead in the desktop publishing market, many of the computer communications products it just not introducing have been available for IBM PC’s for several years.” (Of course, Microsoft was no stranger to long-delayed releases of what turned out to be underwhelming technology, such as the OS/2 Presentation Manager that Tim Bajarin referred to in the future tense in this episode.)

Markoff also said that some analysts and vendors were unhappy with the “authoritarian” approach Apple decided to go with by having a single file server as opposed to adopting a more decentralized approach to networking. We saw some examples of low-cost, decentralized networks in the previous Chronicles episode focusing on the PC platform. And in fact there was an October 1985 episode where Nat Goldhaber of Centram Systems West demonstrated his company’s networking product, TOPS, which made it possible to create a distributed file-server environment to network both Macs and PCs. Goldhaber was understandably miffed that Apple was now marketing a direct competitor to his product and he told Markoff it was somewhat ironic given that Apple originally contracted him to develop TOPS before deciding to pursue AppleShare.

It’s also worth noting that AppleShare was far removed from Apple’s “low-cost” LAN dreams of 1983. The AppleShare software was $799, which actually was quite reasonable for the time. But you also needed to purchase a separate Macintosh with a hard disk to run as the file server, which you didn’t need to do with something like TOPS. Even if you purchased the lowest-end machine capable of running AppleShare–a Macintosh Plus–that would have run about $3,200 with the hard drive in mid-1987. And that’s before you factored in the per-node costs and any other add-ons, such as the EtherTalk card, which originally retailed for around $700.

Notes from the Random Access File

  • This episode is available at the Internet Archive and first aired during the week of November 25, 1987. The recording on the Archive is a rerun from the week of March 17, 1988.
  • Paul Schindler reviewed Webster’s Electronic Thesaurus (Proximity Technology, $90), which is exactly what you think it is.
  • Arthur Young & Co. merged with Ernst & Whitney in May 1989 to form Ernst & Young. At the time of the merger, there were eight major professional service firms. Today there are just four, including Ernst & Young.
  • Bill Yundt worked for Stanford University for 25 years. While still at Stanford in 1989, Yundt took over as CEO of BARRnet, one of the earliest Internet access providers in the San Francisco area. Stanford sold BARRnet to BBN Corporation in 1994. Barr stayed on as a vice president with BBN until 1996, when he joined WebTV Networks, which was later bought by Microsoft and turned into MSN TV. Barr remained with Microsoft until his retirement in 2001.
  • Yundt referred to Stanford creating software to join their computer networks using TCP/IP, which is the foundational protocol for the Internet. The software was originally developed by William Yeager, a research engineer at the Stanford medical school, according to a 2002 article by Pete Carey of Knight-Ridder. As the software became the campus standard, another group of Stanford staffers decided to develop their own routers incorporating Yeager’s work. These staffers would leave the university and form Cisco Systems, Inc. Carey said that Stanford officials contemplated legal action to stop Cisco as they had effectively stolen the university’s IP. Yundt told Carey he favored licensing the software to Cisco instead, as Stanford couldn’t enter the commercial router market anyways, and that was what the university ultimately did in April 1987.
  • Over time, the AppleTalk protocol was superseded on the Macintosh by the TCP/IP standard as the Internet came into its own during the 1990s. But Apple continued to develop AppleTalk up through the release of the first OS X in 2001, and it retained legacy support through the release of version 10.6 in 2009.