Why all the push for the “desktop metaphor” in personal computing? Where is it going and why should you care?

Reprinted from Personal Computing, issue 8/1984, pp. 64-75.

Magazine cover

It is likely that by the time you read this issue, certain hardware and software companies will be flooding the marketplace with claims designed to leave no doubt that: A) Something called the “desktop metaphor” is absolutely essential to increased productivity; and B) that their proprietary implementation of this somewhat amorphous generic concept is exactly the one that’s right for you. This article will tell you what the new desktop environment is really all about, where it came from and how it may change the way we think about computing.

In its simplest, most basic manifestation, says Bill Coleman, director of product development for VisiCorp, “The desktop metaphor is the apparent interaction of several applications (programs) through overlapping windows on the (computer’s) screen.” In other words, where your computer once had to use its entire screen for displaying a spreadsheet or a word processing document or a data base form, with the new desktop software you can simultaneously open several “windows” on your monitor. One can then be devoted to word processing, one to a spreadsheet, one to a data base and so on.

At that point, the logic goes, the screen “becomes,” in effect, a typical desk top on which several documents or work folders might be arranged so that they overlap one another. But the significant technical implication of all this is that each of the applications you display in a screen window is present in RAM or otherwise available at a moment’s notice. Work proceeds through several distinctly different operations without the customary interruptions for reloading of programs or reformatting of data – just as work would have proceeded in the former paper desk top environment, through simply reaching out for whatever was to be done next. But there are some subtleties and even a few complications.

As Coleman notes, “Windows in and of themselves don’t do much at all.” In fact, if you’re only using one application, having several windows open on the screen can be counterproductive, since each window shows a much smaller portion of a document or spreadsheet than could be displayed on the entire screen. Thus, most windowing software allows you to “zoom” in on one window and hide the others, giving the entire screen over to the application of your choice.

That being the case, what good are windows in the first place? As Coleman and others point out, windows are really doors that allow the computer to become more than a single-application machine.

Jeff Elpren, former president of Simtec (a sales organization that proved to be one of the most successful dealers of Apple’s Lisa, one of the first windowing computers), believes that windows are an essential step in the evolution of the personal computer. “There’s a tremendous need to be able to work on an interrupt basis,” says Elpren. “That’s how humans work. They get started on one project, get interrupted and then have to spend five minutes on another project. This has been the shortfall of micros all along. All your wonderful things like your time scheduler and notebooks of all your numbers and names – all these things that look like great applications in fact don’t work. Why? Because to use them for 30 seconds you’ve got to get out of what you’re doing and boot them. In that respect windowing really makes a lot of sense. It makes some of those applications viable.”

From whence it came

Although windowing systems are only now corning to the attention of the general public, the desktop metaphor made its first commercial appearance in 1981, in the form of the Xerox 8010 Star Information System. At a time when managers were just starting to accept the idea of using a simple spreadsheet on an 8-bit computer, the Star appeared to be nothing less than a blueprint for the future. It featured a high-resolution, bit-mapped screen on which files and programs were represented by icons (for instance, small pictures of file folders rather than a list of file names) and presented through windows. All commands were executed, all files and programs were selected by using a mouse to move the cursor, rather than through keystroke sequences. Multiple windows could be opened on the screen and data could be passed “through” the windows, between applications programs, with ease.

Designed to work either on a standalone basis or as part of an integrated office environment, the Star could be equipped with 512k of RAM, 10 Mbytes or more of hard disk storage and an Ethernet local area network with electronic mail facilities.

These capabilities were the result of a long development process which began, according to John F. Shoch, president of Xerox’s office services division, in the early 1970s. “We made the commitment to a number of workstations, including, in particular, those that did the very sophisticated desktop metaphors – graphics, icons and so on,” Shoch says. The purpose of these efforts, according to Shoch, was to design a computer that could be used not only by the hobbyist or technician, but by the untrained person who wanted to be able to work on a computer without ever having to think about how a computer works.

Among the building blocks for the Star were Xerox’s experimental Alto computer, on which early versions of the desktop metaphor system with Xerox’s Ethernet local area network were developed, and the developmental Smalltalk programming language. The latter was a medium in which, Shoch recalls, the engineers at Xerox’s Palo Alto Research Center (PARC) “spent a lot of time on user interfaces and mixing graphics and text.” In addition, an on-line system developed at Stanford Research Institute by a group headed by Doug Englebart, “used a mouse as a pointing device and had a very high quality text presentation,” according to Schoch.

Xerox took a unique approach toward system design in putting those pieces together to produce the Star. As Jonathan Seybold noted in his 1981 review of the Star in The Seybold Report (Vol. 10, No. 16), “Most system design efforts start with hardware specifications, follow this with a set of functional specifications for the software, then try to figure out a logical user interface and command structure. The Star project started the other way around – the paramount concern was to define a conceptual model of how the user would relate to the system. Hardware and software followed from this.”

Despite its state of the art technology, the Star’s market impact has been anything but overwhelming. Part of the reason for this is price. Although the Star has benefited from recents drops in the cost of technology, so that a stand-alone unit can now be purchased for $8995 (which Shoch notes is “within shouting distance of a fully-equipped IBM XT”), the bulk of Xerox’s installed base of Stars sold at a price point over $15,000 – which effectively put it out of the reach of most personal computer users. As a result, the popularization of the desktop metaphor did not occur overnight. In fact, most of the public thought they were seeing something entirely new in 1983 when Apple introduced Lisa, with its bitmapped graphics display, screen icons, multiple windows and mouse controller.

Apple’s Bill Atkinson, however, readily admits that he and his colleagues on the Lisa project were heavily influenced by the work done at Xerox PARC. “Early in the game (circa 1979) we took a look at Alto,” Atkinson recalls. “Alto was our first sight of the desktop metaphor. We saw it for about 90 minutes total, so we certainly didn’t learn how to implement it there. But we got an idea that overlapping sheets of paper were a decent metaphor to express an environment where you really do manage many tasks.”

The year of the metaphor

Like the Star, Lisa was widely acclaimed by computer industry representatives, but widely ignored by the computer-buying public. Again, price probably played a large role in the public’s disinterest, since Lisa (with its bundled software), was released at a price of $9995.

Thus, it remains for 1984 to go down in history either as the year in which the desktop metaphor truly becomes a concept for mass installation – or else the personal computing industry’s folly of all time. Already this year we’ve seen the release of the Macintosh and the new, more powerful and lower-priced Lisa2 from Apple. The desktop environment’s software parade has been joined by VisiOn, Quarterdeck DesQ, Microsoft Windows, Symphony (Lotus Development Corp.), Framework (Ashton-Tate), Enable (The Software Group), Concurrent PC-DOS (Digital Research), Core Executive (Application Executive Corp.) and a host of other packages. These products are designed to make windowing software affordable and to move it to the most popular personal computers.

Robert Carr, chairman of Forefront Corp. (which developed Framework for Ashton-Tate) says, “Like the spreadsheet, the desktop metaphor is almost a standard. It’s a standard because it works. If you look at just the desktop aspects, I think you’ll find that all the systems are very similar. VisiOn, Lisa, Framework and the Star are all much more similar than they are different when it comes to the desktop metaphor.”

Why has this type of software suddenly become so popular and seemingly so important? “There are two main driving forces behind the desktop metaphor,” Carr says. “One is the need and urge to somehow expand that limited screen area to give the user more functionality and viewing capability. The other is the need to present a better system for putting information in order.”

Apple’s Atkinson, who played a major role in designing the user interfaces for both Lisa and Macintosh, adds that, “Windowing is not just expanding the screen width. It really is more a function of retaining context. If you’re in the middle of a document and you have to get out a calculator to do some calculations and then put back the calculator, without windowing you’re talking one screen going away and the new screen with the calculator coming up. You have no reminder of where you were. But when you see a little piece of the document window sticking out there under the calculator it gives you a feeling of continuity and a reminder of context. You have a feeling that the document is still there ready for you, waiting to be worked on.”

Ideally, VisiCorp’s Coleman believes, the end result of all this should be that, “Once a user has learned how one application operates it should be obvious how he works with other applications. And moving data or viewing his data differently through different applications should be a natural extension without going through a lot of reloading programs and saving files and all that. What we’re really saying is the user shouldn’t have to think about how to use the system in order to use it, he should only have to think about how to solve the problem.”

It remains to be seen whether the software buying public will embrace these products, since they are only now becoming widely available. The answer would seem to rest on a number of factors, including how strong the “need to work on an interrupt basis” really is, and how well these packages perform in meeting that need.

How fast is friendly?

One question that arises in many people’s minds is, doesn’t all this attention to ease of use slow down the experienced user. In fact, doesn’t ease of use imply mediocre performance?

Bob Hamilton, vice president of software development for The Software Group, argues that it doesn’t, saying that performance should be measured “not so much by what the computer can do but by what the person at the keyboard can do. You’re measuring performance as applied to the human in terms of their work environment, not necessarily how fast bits and bytes can move behind the screen. The target is to improve the human performance. It’s a different perspective, because early in the game it was just trying to get an order entry system or payroll to work within a certain period of time. When you say performance to people who have been around computing for a long time, they immediately think of what the performance is relative to the hardware as opposed to the tools that are provided to the user.”

In addition, most developers of desktop software seem to recognize that the speed with which a system responds is one of the measures of ease of use. As Apple’s Atkinson says, “When you come down to it, what’s the most user-unfriendly thing you can do? It’s to be slow.” Hal Stegar, Digital Research’s product manager for Concurrent PC-DOS, adds, “One of the worst things you can do to people is to make them wait.”

All these developers claim that performance speed was at or near the top of their specification list in designing these systems. Nevertheless, there is legitimate doubt in many cases as to how well they’ve met the performance specifications. Apple’s Macintosh, for instance, has been criticized for frequently making the user wait an inordinate period of time while it performs various disk functions behind the scenes. Forefront’s Greg Stikeleather, describing the effect of delays like this, says, “Picture turning the knob on your stereo and having to wait 10 seconds for the volume to go up. If you start turning several knobs before you wait for the reaction, the amount of time it takes to learn what happens and the annoyance in waiting becomes very high.” On a computer, he adds, “Those sort of pauses are short enough so that you can’t do anything else useful, but long enough to break up your train of thought.”

The bottom line is that no matter how friendly a system is, it has to perform well. Elpren asserts that even with a system billed as being the ultimate in ease of use, such as Lisa, performance ended up being the bottom line. “The heavy users were the people buying Lisas, not the newcomers or light users,” Elpren says. “It was an applications sale. The people who were buying it were usually looking for some very powerful facilities that they could only get with that environment. Specifically, the two that really stuck out were LisaDraw, with its exceptional graphics, and LisaProject. To sell it, we went after corporate departments that had project management going on.”

At the same time, Elpren notes, the original version of Lisa was undermined by the weakness of some of its other applications. “LisaCalc wasn’t competitive with Lotus 1-2-3,” he says. “Even just the process of moving the data from the spreadsheet to the ‘clipboard’ and back into LisaGraph was a little cumbersome, especially if you had to do it over and over.”

Lotus 1-2-3’s advantages were not coincidental. Lotus president Mitch Kapor recalls that in designing 1-2-3, and later Symphony, “We had the notion of a single unitary data structure. That is, to do a graph you just hit the graph key. If you change a number and then hit the graph again there’s the new graph instantly. The advantage of that approach is that you don’t get into the kind of cutting and pasting operation that you would otherwise have to do.”

The differences

Beyond the presence of windows, the differences between various desktop software packages multiply rapidly. All of these packages are designed in one way or another to make it easier for the user to work with multiple applications and to move data between applications. But developers have taken a number of different approaches in the process of implementing these design goals.

Among the key questions raised by these different approaches are: Are windows enough or are the icons and mouse control – which many of the new systems ignore – necessary elements of the true desktop environment? Can these systems be used effectively to integrate off-the-shelf software, or are customized applications packages necessary? Is true concurrent operation of various applications necessary or merely the appearance of concurrency?

On the Star and Lisa, and later the Macintosh, windowing was tied to an icon and mouse-based command structure. There are no Control or Alt keys to be found on these machines – to select a command you use the mouse to move the cursor to an icon or a pull-down menu. But as desktop environment software migrates to the IBM Personal Computer, very few of the new software packages use the mouse and fewer still use icons. There seems to be some doubt as to the usefulness of the mouse in day-to-day activities such as word processing, and there is a general feeling that icons are not viable with the limited graphics resolution of the IBM.

Kapor says that “the appropriate use of icons and the standard user interface such as on the Macintosh really does simplify things to the extent that you can present the same functionality that you otherwise would and make it more accessible and easier to use.” Nevertheless, while Lotus’s product for the Macintosh will support icons and the mouse, neither 1-2-3 nor Symphony does so on the IBM, a decision Kapor says was based on finding “the best fit for the technology.” Chris Morgan, Lotus’s vice-president of communications, adds, “The metaphor has to make sense with the given hardware. We found that most people who own IBMs don’t own a mouse and we didn’t want to get into the hardware business.”

The question is, how well does the desktop metaphor work without an icon-based command structure. Elpren argues that icon-mouse technology is very important. “It is intuitively right. Apple’s execution of it may or may not be the right one, but it does get us to a higher level of communication,” he says. “The history of computer science for the last 25 years has been nothing except that very concept of, step-by-step, getting to a higher level of communication... from Assembly language to higher programming languages to higher level languages like the data base languages. I see the icons on the desktop environment as just part of that process. They let you communicate with the machine very fast by just ordering them. It doesn’t matter whether you do it with a mouse or touchscreen-whatever.”

Others argue that in many cases the user might be better off without either the mouse or the icons. Charles Mauro, president of a New York-based human factors engineering and design firm, says, “When you combine the mouse and the icons you have a new language, a language in which there are no verbs. A good analogy is an Indian language, Hopi, which has no verbs. If you want to go down to the water you have to say ‘down water’ and imply the direction with a hand motion. The same thing happens with a mouse. If you want to open a file you position your hand on the mouse and you move it and the arrow moves to the file. You click on that and that’s the action. So all the verbs take place by hand action. That’s very appropriate for gross positioning or very simple tasks, but as soon as you get into anything that’s more complex than that you’ve simply got to revert back to a menu-driven interface.

“I think the underlying concept (of the mouse-icon interface) is correct,” Mauro adds. “The thing I found objectionable is that it’s positioned as a panacea. Manufacturers equate ease of use with a mouse. The mouse is only appropriate for a very small number of specific tasks. It’s excellent for manipulating predrawn or rectilinear objects on an established grid, but it’s basically useless for freehand drawing. Cursor keys are easier to use in word processing.”

VisiCorp apparently made similar observations in developing VisiOn. “We designed it from the start with the mouse,” Coleman recalls. “Early on we were very religious about it. We lightened that up a fair amount before we shipped. We’re still moving more and more in that direction to allow things to be done without a mouse.” Other companies have made the same compromise. With Quarterdeck DesQ, for instance, you either issue commands through the time-honored control-key method or use a mouse to activate pull-down menus and select commands from them.

One of the most distinguishing characteristics among the various desktop environment packages now appearing is their degree of open endedness. Some packages, such as Symphony and Framework, combine windowing functions and applications in one large program. Others, such as VisiOn (and the Macintosh, Lisa and Star computers), provide an operating environment that requires customization of software intended to run within it. Still others, such as Quarterdeck DesQ, Concurrent PC-DOS and Core Executive, claim to bring the features of a windowing environment to any MS-DOS program. Each of these approaches has its own advantages and disadvantages. The all-in-one packages, such as Symphony and Framework, are fast because all their applications are combined in one program. Equally important, they can provide a higher level of data integration than other programs. For instance, a graph can be defined as dependent upon the range of a spreadsheet, rather than upon the virtual numbers that reside in that range at the time you define the graph. Thus, if the numbers change, the graph is automatically updated.

The problem with the all-in-one approach is that the applications they provide may not provide all the depth of the best stand-alone applications and you are usually limited to the applications the developers choose to provide, chiefly such common applications as word processing, graphics, spreadsheeting, data base management and communications. (In some cases, however, the developers of all-in-one packages have made provisions for the incorporation of additional functions within the original package. One example of this is Lotus Development’s Symphony, which is designed to incorporate specially developed “add-on applications.”)

Products such as VisiOn, Lisa, Macintosh and Star provide what is called an “operating environment” within which specific application programs can be used. The advantage of this sort of environment, as Coleman explains, is that, “You can extend them simply by installing new programs. They provide several levels of data integration and if there are some standards in the system they can provide integration of the user interface, so that performing a specific function in one program is always the same as doing it in any program.”

A consistent command structure is an important factor in the usability of any system. Without it, you have what Mauro calls “a negative transfer.” Negative transfer is the worst kind of human factors engineering problem you can have, Mauro says, “because if you have a command that does one thing in one program and behaves slightly differently in another program, the user in effect has to unlearn and relearn that everytime he goes through that process.”

The disadvantage of operating environments is that they, like the all-in-one integrated packages, suffer from limited expandability. As Paul Hunt, president of Application Executive Corp., notes, “An environment like VisiOn requires new programs to be written in order to run on it and it requires a sizable investment on the part of the developer for the tools to write those new programs. What has been written has generally been in the tone of generic packages, like a spreadsheet package, a graphics package, a text processing package. But the personal computer is used much more widely than that. People need accounting systems, inventory control, order entry, all those things, or they have specific vertical market packages that they need.”

This problem is compounded by the growing proliferation of these environments. If a clear leader or an accepted standard emerges in this field-on the order of the emergence of MS-DOS as an operating system – then it would be reasonable to expect that a large number of software developers would invest the time and funds necessary to develop widespread applications to run under the winner’s environment. But as long as a slugfest goes on between VisiOn and Macintosh, and Microsoft Windows and others, the end-user of these systems may be limited to generic applications.

Thus, with Core Executive, Hunt’s firm has taken an approach similar to that used in Quarterdeck DesQ and Digital Research’s Concurrent PC-DOS, by allowing the user to load multiple off-the-shelf MS-DOS programs into memory and by providing windowing and data transfer functions for use with those programs. “The availability of good software in a market like the one for the IBM Personal Computer is just fantastic,” Hunt says. “So the idea is to take those concepts – to be able to use programs like you would in a desktop environment – and extend it to the existing market of software.”

The shortcoming of the general integrator approach is that the level of integration they provide is rather minimal. They give you the ability to create multiple screen windows and to transfer ASCII data from one window to another, not the ability to provide dynamic data links or to include, for instance, a Lotus 1-2-3 graph in a WordStar document. Plus, they don’t provide a great deal of ease of use. To run a spreadsheet in one window, a word processor in another and a communications package in a third with these environments, you still have to learn the different command structures of each of those programs as well as the command structure of the environment that integrates them.

The last big point of difference between these programs and machines is the question of concurrent operation or multitasking of multiple applications. In a system such as Lisa, for instance, the user can actually have several programs running at once in different windows. Concurrent PC-DOS and Core Executive both offer a similar capability on the IBM.

Hal Stegar, product manager for Digital Research’s Concurrent PC-DOS, explains the advantages of concurrent operation. “Say you have dBASE II. It could take 15 minutes to an hour to sort a dBASE II file,” Stegar says. “With concurrency, you could get that sorting started and then go to another window and work on 1-2-3 or WordStar.

“Communications is really one of the main uses of concurrency,” Stegar says. “Right now the only way to do communications is to have a personal computer that’s dedicated solely to it. With Concurrent PC-DOS, you can do your communications in the background and your other work in the foreground.” Hunt agrees. “Terminal emulation is an ideal application for our product,” he says. “You can have one of your applications be a terminal emulation package that’s communicating with a host mainframe. So you can in effect integrate a timeshare program in the background while you run something else in the foreground.”

Since the IBM Personal Computer wasn’t really built for multitasking, there is some performance degradation – slowing of applications – associated with the packages that use it. “Performance degradation was one of my biggest concerns when we were designing the product,” Stegar says. “There would be some degradation if you were sorting four dBASE II files at once, because they’re all CPU intensive. Likewise if you were running four copies of WordStar at once there would be some degradation because that’s all screen input and output. But you can sort a dBASE II file and type into WordStar without any noticeable degradation of typing speed.

Where is it going?

Forefront’s Carr says, “I don’t know if the desktop metaphor itself is going to evolve further, but I think the needs that caused it to be invented are still at work and will cause more things to be invented.”

Those needs consist primarily of making it easier for the non-technician to take advantage of a computer’s power. Today that power is largely limited to the area of the individual user’s desktop, but industry leaders seem convinced that in the future it will be extended. “I really think that the true usability of systems like VisiOn and the Star environment and the Lisa/Macintosh environment is going to come in a year or so when you’ve got a little more powerful processors and you’ve got local area networks,” Visi-Corp’s Coleman says. “It’s really going to be based primarily on communications: local area networks, disk servers, file servers, electronic mail and messaging, as well as a distributed data base and gateways into other in-house data processing functions. Once that stuff all comes to light you’re really going to see the need for these things that are much better integrated.”

How well the new desktop environment works and how well it sells will likely be determined before that happens. Its progenitors will tell you today that the single-function machine is a thing of the past, virtually obsolete. But time will tell which of the varied, and conflicting, approaches to the desktop environment will be embraced by the software-buying market, and which will be deemed inadequate or merely superficial. We can confidently predict there will be some of both.

by Paul Bonner (Senior Editor)

Sidebar:

	“Desktop vs. IBM and clones”