32by32 - Macintosh History from the 1980s

Cayman GatorBox

“Nothing else currently does this”

This is a story of a very unique networking product, the Cayman GatorBox. A flawed device, to be sure — but also a compelling synthesis of many different file sharing protocols and communication languages, integrated into one small metal box. For just a few thousand 1980’s dollars, Cayman promised to link together Unix workstations on DARPA-class Ethernet with Macintosh computers connected by Apple’s small-scale, friendly AppleTalk cables.

An intelligent gateway

Our journey starts in January 1988. at the Infonetics Desktop Communications Conference. It was at this Anaheim event that Cayman Systems Inc. of Cambridge, Mass. announced GatorBox, an “intelligent LocalTalk-to-Ethemet gateway” that “lets Mac users operate AppleTalk applications on Ethernet”, in the words of company President Ted Stabler. The GatorBox was due to ship in April 1988, at a price “competitive” with other products.

By March, MacWEEK hand nailed down that price: $2,295. The May issue of Macworld was written before that price was publicized, but did contain details about the “network interfaces for LocalTalk network and Ethernet or ThinEthernet”. The first of these was what we call Thick Ethernet today, expressed as a 15-pin AUI connector, while the second was the coaxial cable for 10Base2. Let’s take a look at those interfaces:

By July, MacWEEK hit the presses with both a delayed shipping date for the GatorBox — August — and a much higher price, $3,495. This price might have reflected the addition of promised future software elements, such as SMTP-to-Microsoft Mail, although that add-on wouldn’t ship for years into the future:

Cayman’s gateway software translates Microsoft Mail messages into a format readable by mail programs based on SMTP, or Simple Mail Transfer Protocol, the electronic-mail transfer protocol used in TCP/IP networks environments.

Looking past this SMTP vaporware, the value proposition for the GatorBox was always a unique blend of hardware and software features. Unlike the simpler and cheaper FastPath from Shiva, Cayman’s product bridged not only physical layers and protocol stacks but user-facing, “Level 7” application protocols such as AFP and NFS.

The GatorBox from Cayman Systems of Cambridge, Mass., provides a bridge between AFP and NFS systems, allowing the two architectures to connect while retaining transparent file access for all users. Using this approach, a Mac user can access Sun files and vice versa. Apple has traditionally looked to third-part vendors to round out its product offerings, and Cayman has been only too happy to oblige.

Handling these protocols directly at the firmware level could mean much simpler client configuration — rather than install an NFS client on the Mac, and compile CAP for Unix machines, the machines retained their native software experience and all translation took place on the GatorBox. In the words of MacUser, “[GatorBox] transparently connects LocalTalk to Ethernet – no software needs to be installed on any machine.” GatorBox soon found customers in higher education, where mixed networks of Macs and Unix machines were common:

At Harvard University, Cayman’s GatorBoxes are being used by 100 students in the Graduate School of Design for information sharing among Macintoshes and Sun workstations. “The NFS functionality of GatorBox has allowed students to download their homework on either Macs or Suns,” said David Kovar, who is a technical consultant for Harvard’s Office of Information Technology. “That’s especially important here since we already have a lot of Suns and just started using the Macintosh last fall.”

Even the presidents of other networking companies were impressed. Writing in the May 1980 issue of MacUser, Farallon head honcho Reese Jones noted:

While Apple and Sun duke it out over protocols, with the use of a Gator Box from Cayman Systems, you can convert AFP to NFS, and LocalTalk to TCP/IP Ethernet. AFP clients on the AppleTalk side can “see” an NFS server as an AFP server, gaining transparency between machine types, alternate file systems, and network operating systems.

Little wonder that president Stabler called his product “the first intelligent gateway from the Mac to Ethernet” — and capital markets agreed. Cayman raised US$1.8M from venture capital firm Concord Partners in early 1989.

And new markets for Unix-to-Mac connectivity were emerging every day: what MacWEEK described as the “the unveiling of the first Mac-to-NeXT connectivity product” took place at Macworld Expo in San Francisco that month. Stabler’s claim that GatorBox had shipped in conjunction with NeXT’s launch in late 1988 suggests that either the GatorBox had slipped from its promised August shipping date — or that a software revision optimized for NeXT’s implementation of NFS had been released in that same time. Regardless, Cayman “…have the opportunity to be at the forefront of an emerging computing frontier… this is as exciting as the Mac’s development five years ago.” Indeed, integrating NeXT and Mac systems wouldn’t get easier until the release of NeXT’s AppleTalk stack in NeXTStep 3 in late 1992, leaving a substantial window for Cayman to sell into workplaces with both of Jobs’ creations.

1989 also brought changes to the fundamental protocol the GatorBox dealt with: AppleTalk 2.0 promised to break free of the node-number and architectural limits that hindered complex integrations. But in addition, it promised software routing between different network interfaces in the same Mac. A Macintosh II with an Ethernet NuBus card could theoretically perform some of the same tasks as the GatorBox itself. “Before, I had to rely on hardware from Kinetics and Cayman”, said one beta tester. With the complete GatorBox setup still retailing for $3,500, probably more than one customer started considering such an option. But Cayman needn’t have worried — others “do not believe that software bridges can provide the functionality or performance of hardware-based solutions.”

Cayman’s response was ready by June of that year: they would “introduce two new products and restructure another in support of AppleTalk Phase 2.” The modular hardware architecture shown in the photo above allowed the release of the $3,500 GatorBox T, with Token Ring taking the place of the Ethernet module. And $700 GatorCard — actually a rebrand of the Racal Interlan — would add Ethernet to the Macintosh II and SE, offering a single-vendor solution for workplaces introducing higher speeds to their Macintosh networks.

But it was the GatorBox itself that saw the most changes in response to AppleTalk Phase 2 — at least in terms of bundling and marketing. While a short-term maintenance update (at a “nominal charge” ) would upgrade the GatorBox’s AppleTalk routing to Phase II compliance, an “extensive restructuring” of the product would see the higher-level protocol translations broken up into optional modules, or “GatorBox applications”. Each module could be downloaded onto a base GatorBox, adding functionality at the same time it added revenue to Cayman’s bottom line.

The new lineup looked like this:

GatorBox: $2,795, AppleTalk-TCP/IP routing only
GatorShare: $1,995, AFP/NFS, third quarter of 1989
GatorPrint: $595, PAP/LPR, fourth quarter of 1989
GatorMail: $995 for 10 Mac users, SMTP/Microsoft Mail 2.0, February 1990

The cost savings were clear — although so was the eye-watering cost of nearly $5,000 to achieve the functionality previously offered for $3,500. The press generally repeated Cayman’s company line:

Though previously sold as one product, Cayman unbundled the two products to appeal both to users who need basic LocalTalk-to-Ethernet routing and to those who need the full functionality of an NFS (Network File System) gateway.

In the new, “unbundled” world, software add-ons could be announced (and priced) as quickly as Cayman wanted market attention (or additional revenue.) Yet the actual software itself often seemed to lag. If journalists noted that GatorMail was “originally announced in July 1988 and now due sometime in early 1990” , Cayman pushed ahead with GatorMail-M, which was licensed from StarNine and offered support for DOS MS Mail clients in addition to Macs. (Although scribes noted Cayman “has not yet set pricing or a release date for GatorMail-M.”) . Next to be announced was GatorMail-Q, a “QuickMail-to-Unix mail gateway based on StarNine Technologies’ Mail*Link [SMTP] software”. . The two StarNine-licensed products ran on the mail servers, no special firmware needed to be loaded onto the GatorBox itself. They finally reached users in February 1990.

GatorMail is transparent to users, according to Cayman. Incoming messages from Unix machines appear under the QuickMail or Microsoft Mail interface. For outgoing mail, long Unix addresses can be hidden by aliasing, so that only the name or nickname of the Unix recipient appears in the directory.

As 1990 dawned, Cayman seemed to be cresting a wave. The GatorBox was demonstrated at trade shows such as Uniforum and at Apple’s own AppleTalk roadshow seminars . Splashy full-page ads caught readers’ eyes in MacUser and Macworld:

…while MacUser devoted an entire four-page article to the GatorBox . If Vernon Keenan’s Sharing: From Apples to Alligators was critical of Cayman’s documentation, it also suggested that the need to bridge Mac & Unix network protocols was mainstream enough to merit coverage in a consumer magazine.

It’s Keenan’s article, in fact, which raised the first contemporaneous criticism I’ve seen of a weakness in the GatorBox product: the need to load the operating system and configuration at every power-on. This limitation was particularly acute in Cayman’s 1.4.1 software release, which required a LocalTalk-only Mac, or a Unix TFTP server, to start up:

The GatorBox’s essential software. GatorKeeper, is stored in a folder on the administrator Mac and is automatically downloaded when the GatorBox is rebooted; the Mac must be turned on, connected to the GatorBox’s LocalTalk network, and have GatorKeeper running to reboot successfully. A UNIX host can also act as a boot server for the GatorBox. This booting scheme is unfortunate, because it relies on the boot server’s being available at all times. It would be better to have the software download to the GatorBox once and remain there, battery-backed, until the next software upgrade. Shiva’s EtherGate is an excellent example of this type of approach.

There’s no better way to describe this flaw than the pointed critique that Keenan offered way back in 1990. indeed, for retro collectors, the original GatorBox’s dependency on an external server to boot nearly rendered all the devices premature e-waste, until the software was located, archived, and shared.

One of the most useful parts of this MacUser test is the real-world performance numbers — and these numbers belie the idea that a powerful UNIX system would translate into a fast AppleShare file server:

Copying a 2-megabyte file from an AppleShare server on EtherTalk to a Mac client on EtherTalk takes 31.5 seconds. It takes 61.5 seconds to copy the same file from the NeXT NFS server to the EtherTalk Macintosh client. The additional overhead comes from the GatorShare gateway.

This stands in contrast the promise that “With the GatorBox, Macintosh users can turn the [NeXT] cube into a high-performance AppleShare server, according to Cayman officials.” Yet the results weren’t consistent across all interfaces — bizarrely, LocalTalk-only Macs enjoyed a speedup that Ethernet-equipped workstations did not:

“When a Mac on LocalTalk communicates with an NFS server on Ethernet, the story changes… GatorShare’s gateway (not GatorBox’s router runction, as one might expect) is used… Seemingly counterintuitive, routing between LocalTalk and EtherTalk with the GatorBox is slower than GatorShare’s translation of LocalTalk AFP to Ethernet NFS.”

Another criticism Keenan had was the ersatz Finder-esque interface that GatorKeeper used to manage devices. “GatorKeeper’s interface, although iconic, is not very intuitive. It’s also nonstandard; the Trash appears in the window instead of on the desktop.” This is a common critique of those Macintosh interfaces which are inspired by the Finder, going so far as to replicate the design language and logic of an interface that the programmers no doubt assumed the user would be familiar with already. But problems are likely to arise when software looks like something expected — and then behaves differently. HyperCard is the best-known example of this — a world unto itself, despite the shared programmers and designers (Atkinson, Kare) with the Macintosh’s own operating system. I’ll reserve judgement until I get my own GatorBox working, as I’ll need to use this software myself.

Yet for all these flaws, the essential promise of the GatorBox shone through: users could finally link their the Sun or NeXT workstations on Ethernet with the Macintosh computers that were increasingly being used for desktop publishing and other graphical tasks:

Nothing else currently does this, so if you want to get at NFS servers through the familiar AppleShare software interface, GatorShare is the product to buy.

Another comprehensive review landed in the June 1990 issue of MacUser.

Towards the 1.5 release

Some of the problems Keenan encountered would be fixed by a new 1.5 software release, which included features such as SNMP, tunneling AppleTalk through TCP/IP networks without EtherTalk, and security options such as hiding devices from users in other AppleTalk Zones. The release also supported protocols such as RIP, MacIP and atalkad, which “let administrators route, configure and monitor AppleTalk packets on TCP/IP networks more efficiently.”

GatorShare was also revised to version 1.5, which supported Apple’s DOS AppleShare client (bundled with their ISA LocalTalk cards), and “byte-range locking, which permits the use of multi-user databases across the GatorBox gateway” . (Keenan acknowledged this feature was coming soon during his MacUser evaluation.) Re-mapping of Macintosh characters into NFS-compliant strings was another improvement. And a small INIT could do the job of loading the initial software onto the GatorBox, supplementing the previous requirement to have a copy of the GatorKeeper application running on a Mac.

Software | Permalink | No Comments

BMUG Meeting minutes 12/17/87

Acta document

BMUG Meeting 12/17/87

Schedule

Discussion & Announcements 6:30 – 7:00

St. Silicon

Beck-Tech

WordPerfect

Raffle

Discussion/Announcements

All products not yet shipping will be out at Expo.

What is NEXT up to?

Mac+: memory upgrade?

What won’t invalidate the warranty?

Buy 4 120-ns SIMMs: $210/meg

up to 2 megs or 4?

Talk to ComputerWare or M.A.C. or CJS or MacOrchard.

Dove MacSnap will piggyback for Mac+.

2.5 megs is good for HyperCard & MultiFinder & some stuff

Get a Hard Disk first.

Mac II with expanded video

30% slower in 2-bit mode than accelerated SE.

Get an SE unless you need gray levels.

Big screen is useful if you use spreadsheets or WP.

Hard Disks: recommended? reliable?

Insides: MiniScribe, Quantum…

Noise: inside & outside affect it

Outsides: Jasmine, LaCie, DataFrame, CMS, MacBottom

CMS: $585 for 20 megs

Apple used MiniScribe/Rodime in SE; Quantum in II.

M.A.C.: 60 meg CMS (Mac II internal) stopped shipping – reliability problems in software.

Consider: Software, Portability.

Jasmine BackPack 40 is finally shipping.

Jasmine just dropped their credit card surcharge.

Mac the Knife in MacWeek: FullWrite manuals were NOT hijacked on that truck in LA. Manifest.

Equipment for sale: SE & stuff: $4,000 or best offer.

300 copies of Goodman’s book: $30. 849-9477.

CostCo has book for $17 – Central Ave. in Richmond

other HyperCard books: at least 5 others.

BMUG BBS: 4 lines

CompuServe: MACFUN, MACPRO replace MACUS.

Q: New plotter drivers? HP GraphMaster II? All the existing ones don’t work well.

A: Try HP. They’re working on it.

Q: Cricket Draw doesn’t work well?

A: Try various plotting programs.

Q: RSG: pasting stuff into MacDraw 1.9.5: strange error msg.

A: It needs room in a particular segment.

HyperCard; the usual questions.

BMUG Minutes | Permalink | No Comments

PixelPaint 1.0

First 8-bit Painting Program

Perhaps the first 256-color paint program for the Mac, PixelPaint was written by Pixel Resources and published by the hardware vendor SuperMac. It was positioned as a software showcase for the company’s new Spectrum 8-bit color video card for the Mac II and accompanying 19″ monitor, introduced at MacWorld Expo Boston in August 1987.

Excitement built about PixelPaint in the trade press leading up to its release.

…SuperMac Technology of Mountain View, Calif., is preparing a color paint program for the Mac II called PixelPaint.

The context was an article about the flood of graphic design programs that extended Apple’s original MacPaint concept: FullPaint added multiple windows and full-screen editing, while PixelPaint focused on complete support for the 256 colors possible in Apple’s 8-bit color standard.

Introduced the previous summer, PixelPaint was shipping by January 1988 – one of the few pieces of color Mac software then available.

SuperMac’s positioning of this software reflected more their sales goals for their large, expensive monitors than it did the eventual landscape of graphics software. The coverage in MacWeek emphasized “Bit-mapped color presentation graphics” “color transparencies for business presentations,” and “presentations for corporate clients” rather than any kind of personal creative work. But in the early landscape of color graphics circa 1987, virtually the entire workflow had yet to be invented. Journalists discussed the notion of using software such as PixelPaint to “colorize” existing artwork – in a time years before color scanners made that a trivial task.

Software | Permalink | No Comments

SyQuest Drives

Removable Drive

The earliest trace of what would become a juggernaut in late-80s/early-90s Macintosh storage was an article in MacWeek in late November, 1987:

MILPITAS, Calif. — Custom Memory Systems Inc. has announced a December ship date for a removable hard disk drive system for the Macintosh. The $1,795,5.25 half-height drive, the CM-55, has an average access time of 25 milliseconds and a 1-to-l interleave — the same performance offered by many hard disks. It stores data on 44.5-Mbyte removable cartridges that are available from the company for $125 each. The CM-55 takes about 10 seconds to come to speed after a cartridge is inserted and automatically tests each cartridge during the process, according to the company. Custom will include software utilities with the drive that will help users configure their system of disks.[…]

In late 1987 CMS was still the exclusive retail partner for SyQuest’s technology — although that would soon change. The name SyQuest didn’t show up in any early CMS publicity material; the company initially tried to brand the new device and its cartridge “CM-55” (The number referred to the unformatted capacity of each cartridge; in actual use of course the capacity was 44 megabytes and CMS’s first ads show a change to “CM-45.”) Nevertheless, SyQuest would become shorthand for the entire system, with that company’s logo appearing on nearly every cartridge sold.

If CMS was determined to play down the OEM manufacturer in these early days, MacWeek reporter Jiri Weiss did the opposite. He dug up a fascinating pre-history of the company — one that did not exactly inspire confidence. Despite having delivered nearly $50 million worth of earlier drives to the federal government and large corporations, Syquest’s technology suffered from a design flaw which occasionally locked cartridges to only working with the exact drives they were first used in:

The company that developed the technology, Syquest Technology of Fremont, Calif., has had problems with reliability in the past, said Phil Devin, senior industry analyst with Dataquest, a market research firm in San Jose, Calif. The problem has been to design cartridges so that they are interchangeable, Devin said. With data packed in closer than on a floppy, it is easy for the drive to miss Track 0, the disk directory, he said. “Even sealed drives have had problems finding Track 0. When you play with hard-track densities, it becomes very hard to design a reusable pack,” he said. Although Syquest has had return rates of 20 percent on earlier models, the problems were solved several years ago, Steve Katamay, CMS president, said. “I wouldn’t have put the earlier models in my product. The disks are now 100-percent interchangeable.”

Also interesting from this November 1987 MacWeek article is that CMS executives were already anticipating the competition that SyQuest would be measured against — the Mac version of the VHS/Betamax debate: “The closest thing to this is a Bernoulli drive, but we are as far above it as a hard disk is above a floppy.” (Bernoulli drives were manufactured by Utah-based Iomega, whose later product Zip would become the SyQuest drive of the mid-1990s.) CMS’s first full-page ad, shown above, was explicit: “Not Tape… Not Floppy… Not Bernoulli.”

MacWeek predicted customer shipments for the CM-55 in December 1987, for an introductory retail price of $1,500 which would soon ramp up to $1,800 (!) for the drive alone. 44meg (sorry, “55 megabyte (unformatted)” cartridges would be $125 each. In retrospect, these early prices seem astronomical — by the time the original 44mb system had hit its stride and achieved large scale deployment, drives would go for for $200 and cartridges $40.

But what need would drive the success required for a successful transition from niche to necessity? In a word: the laser parlor. An essential (if now vanished) element of the 1980s Desktop Publishing scene, these retail operations offered walk-in access to high-end printing equipment that the average user could never afford. Although as time went on some offered professional-class imagesetters with dedicated outboard RIPs (raster image processors), in the early years many provided nothing more than access to an Apple LaserWriter Plus or IINT. The $5-$8,000 price point of these early laser printers — although inexpensive compared to even a few years previous — still rendered them inaccessible for the average freelance designer or page layout artist. (We are still a few years before the 1989 introduction of the HP DeskWriter and its democratization of pseudo-300dpi output.) So the solution was simple: take your PageMaker or Adobe Illustrator files into a facility where you could print out your work in high resolution, and perhaps have it duplicated at the same time for mass distribution.

In order to do that, you needed to fit your work on an 800k disk. Easy enough for a simple vector illustration, or a text-heavy DTP project. But what about large, multi-page projects with complicated EPS and large scanned TIFFs? And how to assure that the laser parlor would have the precise fonts you designed with — which, keep in mind, were licensed only to you? Juggling multiple floppy disks to dynamically reassemble a complex project, when laser parlors were charging by the hour as well as the page, would be frustrating and expensive. And in the days before a reasonably-priced laptop (the $7,000 Macintosh Portable cost more than a laser printer), bringing in your entire workstation would be a literal heavy lift.

Enter the SyQuest cartridge. With 44 megabytes of space, even the largest and most complex desktop publishing project would fit, with room to spare for the outline and screen fonts, linked external TIFFs, precise version of PageMaker or Quark with all the needed plugins — even your own System Folder, if you wanted to boot directly from the cartridge itself. Years before the NeXT Cube and its magneto-optical drive would become widely available, the SyQuest 44mb system offered the real-world, affordable experience of what Steve Jobs promised in the introduction of that futuristic computer: bringing your entire data world with you in a portable disk.

Unlike the black Cube’s rather over-engineered technology (lasers at two different intensities, to either read the magnetic disc or write to it by heating each bit to the Curie point), the SyQuest system was simple: essentially an externalized magnetic hard drive platter, encased in a rugged semi-translucent plastic rectangle and reasonably sealed to keep out dust while in transit. Upon inserting the cartridge, the drive drew open a black plastic arc of a door, inserting its read/write heads above and below the disk as it entered. A characteristic “clicking” spin-up sound followed, accompanied by a flashing light that cycled from orange to green as the drive became ready. Once loaded, the disk was easily as fast as the average internal or external hard drive in most cases (a refreshing contrast to the magneto-optical drive NeXT used, whose speed was so slow the company began to ship traditional 40mb “caching” drives to bolster the systems’ reaction times.)

SyQuest escaped the orbit of its first reseller, CMS, with a second deal with Mass Micro in January 1988. :

and asserts its own brand as it becomes a must-have peripheral, available from vendors such as MassMicro, Jasmine, Microtech, and others. Meanwhile, the earliest mention I can find of SyQuest in a user context is an August 1988 column from BMUG’s David Morgenstern in MicroTimes. Morgenstern usually wrote a summary of the latest news and gossip drawn from the weekly BMUG meetings, so it’s safe to say that SyQuest had become a big topic of conversation during the summer of 1998:

One of the favorite new toys of the Mac power-users with power-bucks are the 44 meg “removable media” drives. These are hard-disk drives with the hard-disk in a removable plastic cartridge. They are the 10-wheel truck of Floppyville. All of the hardware manufacturers are coming out with some form of these drives. D.P.I, MassMicro and Peripheral Land have been leading the pack with big ad budgets.

Interesting is that David never mentions “SyQuest” brand name in his piece — apparently the large size of the removable storage (44MB) was a better descriptor than the OEM vendor of the actual mechanism.

MacWorld magazine first mentioned SyQuest drives in the context of a Backup article in November 1988. Given print magazines’ lead times — and the complexity of an article that reviewed 50 separate products — it’s safe to say this also reflects a Summer 1988 timeframe for the transformation of SyQuest: from newly-introduced gadget to essential accessory.

Hardware | Permalink | No Comments