Apple Open Collaboration Environment

AOCE/PowerTalk was heavily marketed between 1993 and 1995, but the hardware requirements meant that most users couldn't even install it, let alone use it. Developers were likewise stymied by the complex system, and by users who didn't seem to expect to have to pay for something that was otherwise intended to be "built in". In 1996 Apple Computer quietly dropped their efforts to market AOCE, and the project quickly disappeared.

History

Development of AOCE started in 1989, largely the "pet project" of Gursharan Sidhu, engineering lead at Apple for LaserWriter, AppleShare and related networking products. At the time, during John Sculley's "don't bother me with technology" era, practically any mid-level manager could arrange to have projects funded if they had enough "pull". This was certainly the case for Sidhu, who was well respected within Apple.

The project started by taking a "20,000 foot overview" of existing mail systems, and trying to find common concepts and problems. The key conclusion they came to was that e-mail systems were confused about their own purpose; their key task is to provide a mechanism for store-and-forward delivery of things to places, but existing systems invariably delivered e-mail to people. Compare this with the real-world postal service, which will deliver not only mail, but magazines, packages, large parcels, and even (in one example) building materials to a worksite.

They also found that existing e-mail systems shared a number of common problems. They tended to support plain text mail only, and rarely included any support for non-English characters. Support for mobile users was spotty at best, often relying on 3rd party "hacks" that were of dubious reliability. And they were all, without exception, based on a dedicated e-mail server that was typically complex to set up, and often "overkill" for small installations with only a few people in an office.

And finally none of the existing products could give the user what they really wanted: a single universal mailbox and a single universal address book. At the time savvy users would often have mailboxes on their corporate network, online services such as CompuServe or AppleLink, and perhaps a number of BBS systems as well. Each e-mail system used its own standards for collecting and storing information, forcing users to run multiple clients to access the different services. Although a single-mailbox system could be constructed by administrators with the use of e-mail gateways, these tended to be very expensive and technically challenging to maintain.

AOCE aimed to fix all of these issues at the same time. At "one end" of the system, AOCE focused on the underlying delivery and addressing systems, generalizing the e-mail concept so the system could be used to deliver anything from e-mail to word processor documents to print jobs. Addressing was another issue the market was struggling with, so AOCE would offer a single universal addressing mechanism and address book, one that could support not only people's e-mail addresses, but the addresses of things like printers and fax machines as well.

Remote users would be supported by native AppleTalk Remote Access, Apple's "standard" solution for supporting the AppleTalk protocol over modems. The system understood that users were not always connected and delivery didn't have to happen until both the sender and recipient were online. Even on a LAN this would be valuable, as many people turn off their computers at night and the mail would have to wait until the next morning for delivery. For security over the potentially "open" phone lines, all communications could be secured using RSA encryption and digital signing, even on the local network.

AOCE would normally run peer-to-peer, with user's mail being stored locally on their machine. This allowed them to go mobile, as well as allowing small networks to be set up for zero cost. Users would be able to send documents directly to other users simply by dropping the address on the document, or vice-versa, bypassing a "message" at all -- the document would arrive, by itself, in the other user's mailbox. Delivery of e-mail, or anything else for that matter, would be handled entirely by plug-ins, allowing the user to collect mail from all of their sources and collect it into their single inbox. An optional server could be installed for performance and maintenance needs for those sites that required it.

In other words, AOCE was suffering from the second-system effect, where engineers spend considerable time designing a system that does everything. Invariably these projects fail as the demands are not only incredibly difficult to meet, but often fail to meet real user needs. Often the ideas themselves are good, but buried inside unusable implementations. Users would not learn how unusable until it was released.

By early 1993 the "client side" of AOCE was nearing completion. Apple started a pre-release marketing campaign, telling their larger customers and even 3rd party e-mail vendors that AOCE would soon arrive and change the market completely. The claim was that nothing else would be able to compete with its ease-of-use, power, and lack of maintenance overhead -- all hallmarks of "the Apple way".

AOCE was publically released in September 1993, part of the System 7 Pro bundle that included AppleScript utilities and QuickDraw GX.

In-use

When the product finally shipped after years of hype, users were dismayed to find that there was basically no way they could install it. An install of the various client-side components required a machine with 2.5MB of RAM minimum, and really needed 4MB to run well. This was the maximum available RAM in many Mac systems of the era. Removing unneeded components did little to address this, and AOCE and the other Apple technology du jour, QuickDraw GX, typically could not be run together because of a lack of memory. While newer machines were able to run AOCE more comfortably, as an e-mail system intended to be run on diverse networks of non-homogeneous machines the requirements greatly impaired market acceptance. Simply downloading and installing separate stand-alone client applications for each mail system the user actually had would use considerably less disk space, and had no constant memory footprint.

It was equally clear that while PowerTalk was generally an interesting system, a combination of design features made it frustrating to use in the real world. For instance, the addressing system was so deeply embedded into the core of the system that simply typing in a new address was an ordeal. First the user had to click on a button, select the address type, type it in, and then finally click OK to have it appear in the message. Disk usage was also ridiculous; each message was stored as a separate file, requiring 1k or more of space in an era where 40MB and 80MB disks were still common. The result was that only a few hundred letters would be enough to fill the free space on the drive.

Another annoyance was due to the fact that the Mac OS did not require users to log in, and thus the system could not know who a user was. This meant that documents had to be delivered to a user's machine. This, obviously, did not work well in the case where the user had two or more machines, making the concept of a universal mailbox annoying in practice. Backing up e-mail was likewise almost impossible; the mail was spread out over the network, some of it remote and unaccessable.

Even the remote access was doomed by feature interaction. To ensure that all messages were delivered in a reasonable time on a network where machines might appear and disappear at random (when they are turned on and off), AOCE had a 15-minute timeout in which it tried to deliver pending messages. However, if the user in question was using a dialup connection on a modem, this meant that delivering mail took of 15 minutes on the phone if any of the recipients were not online.

Many of these problems were intended to be solved with the PowerShare server, which acted as an always-on, always responsive "super-peer". The basic AOCE protocol would notice these machines when attempting delivery, and send to them first, thereby eliminating the delays and centralizing storage and maintenance. Sadly the server was not ready in time for the release, and didn't ship for another year. When it did it was likewise slow and resource hungry, largely a side effect of various features of the Mac OS that made it unsuitable for server applications (not that it was designed for this role).

AOCE had one year in the sun at Apple's Worldwide Developers Conference in 1995 when it was finally gelling, but by this point almost everyone realized that the market was clearly moving towards SMTP-based internet mail as an almost universal format. By 1996 Apple had given up on AOCE, and started talking about the CyberDog project based on the OpenDoc platform. AOCE quietly disappeared.

Given reasonable staffing estimates over that time, it is possible that the project cost anywhere from $50 to $100 million at this point, thereby adding $50-100 to the price of every Mac sold.

Description

The system was designed in an era when there were many e-mail formats and services, including online services such as CompuServe and AppleLink, networking standards like X.400 and SMTP (internet mail) and LAN-based servers such as Microsoft Mail and QuickMail. In order to support this diverse environment, AOCE included a robust layered protocol stack that, in theory, could be used with practically any store-and-forward type of environment. This was used within AOCE not only to service mail, but faxes, printing and even directly sending files from one machine to another without enclosing them in a mail message or needing a file server.

Adoption of AOCE among 3rd party developers was slow due to a ferociously complicated API. The book documenting the system was larger than all of the books describing the rest of the pre-System 7 Macintosh put together. Adding a simple feature like "mail this document" to an application required wading through hundreds of pages of documentation, and writing a core AOCE component was many times more complex.

Several parts of the AOCE engine were useful on their own, notably the Keychain. However in order to get the keychain, you had to install all of AOCE, a cost the users were not willing to pay. Many years later the keychain was finally shipped as a stand-alone component in Mac OS 9. The encryption/signing support is arguably also useful (although not widely used in the "real world"), but has not reappeared in a stand-alone version to date.

Other concepts in AOCE remain intriguing and have yet to appear in other systems. For instance the idea of being able to drag a document onto a user's address (which today would be standardized using a VCARD) to directly pass the document on the next time that user appeared on the network seems like an excellent idea. Likewise the concept of peer-to-peer mail is something almost no systems offer, although the need for this in an era of ubiquitous mail servers is arguable.