• Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint
Share this Page URL

Chapter 3. Wasting Money > The Cost of Prototyping

The Cost of Prototyping

Prototyping is programming, and it has the momentum and cost of programming, but the result lacks the resiliency of real code. Software prototypes are scaffolds and have little relation to permanent, maintainable, expandable code—the equivalent of stone walls. Managers, in particular, are loath to discard code that works, even if it is just a prototype. They can't tell the difference between scaffolding and stone walls.

You can write a prototype much faster than a real program. This makes it attractive because it seems so inexpensive, but real programming gives you a reliable program, and prototyping gives you a shaky foundation. Prototypes are experiments made to be thrown out, but few of them ever are. Managers look at the running prototype and ask, “Why can't we just use this?” The answer is too technically complex and too fraught with uncertainty to have sufficient force to dissuade the manager who sees what looks like a way to avoid months of expensive effort.

The essence of good programming is deferred gratification. You put in all of the work up front, and then you reap the rewards later. There are very few tasks that aren't cheaper to do manually. Once written, however, programs can be run a million times with no extra cost. The most expensive program is one that runs once. The cheapest program is the one that runs ten billion times. However, any inappropriate behavior will also be magnified ten billion times. Once out of the realm of little programs, such as the ones you wrote in school, the economics of software take on a strange reversal in which the cheapest programs to own are the ones that are most expensive to write, and the most expensive programs to own are the cheapest to write.

Writing a big program is like making a pile of bricks. The pile is one brick wide and 1,000 bricks tall, with each brick laid right on top of the one preceding it. The tower can reach its full height only if the bricks are placed with great precision on top of one another. Any deviation will cause the bricks above to wobble, topple, and fall. If the 998th brick deviates by a quarter of an inch, the tower can still probably achieve 1,000 bricks, but if the deviation is in the fifth brick, the tower will never get above a couple dozen.

This is very characteristic of software, whose foundations are more sensitive to hacking than the upper levels of code. As any program is constructed, the programmer makes false starts and changes as she goes. Consequently, the program is filled with the scar tissue of changed code. Every program has vestigial functions and stubbed-out facilities. Every program has features and tools whose need was discovered sometime after construction began grafted onto it as afterthoughts. Each one of these scars is like a small deviation in the stack of bricks. Moving a button from one side of a dialog box to the other is like joggling the 998th brick, but changing the code that draws all button-like objects is like joggling the 5th brick. Object-oriented programming and the principles of encapsulation are defensive techniques whose sole purpose is to immunize the program from the effects of scar tissue. In essence, object orientation divides the 1,000-brick tower into 10 100-brick towers.

Good programmers spend enormous amounts of time and energy setting up to write a big program. It can take days just to set up the programming environment, before a line of product code is written. The proper libraries must be selected. The data must be defined. The storage and retrieval subsystems must be analyzed, defined, coded, and tested.

As the programmers proceed into the meat of the construction, they invariably discover mistakes in their planning and flaws in their assumptions. They are then faced with Hobson's choice of whether to spend the time and effort to back up and fix things from the start, or to patch over the problem wherever they are and accept the burden of the new scar tissue—the deviation. Backing up is always very expensive, but that scar tissue ultimately limits the size of the program—the height of the bricks.

Each time a program is modified for a new revision to fix bugs or to add features, scar tissue is added. This is why software must be thrown out and completely rewritten every couple of decades. After a while, the scar tissue becomes too thick to work well anymore.

Prototypes—by their very nature—are programs that are slapped together in a hurry so that the results can be assayed. What the programmer exchanges in order to build the prototype so speedily is the perfect squaring of the bricks. Instead of using the “right” data structures, information is thrown in helter-skelter. Instead of using the “right” algorithms, whatever code fragments happen to be lying around are drafted for service. Prototypes begin life as masses of scar tissue. They can never grow very large.

Some software developers have arrived at the unfortunate conclusion that modern rapid-prototyping tools—such as Visual Basic—are effective design tools. Rather than designing the product, they just whip out an extremely anemic version of it with a visual programming tool. This prototype typically becomes the foundation for the product. This trades away the robustness and life span of the product for an illusory benefit. You can get a better design with pencil and paper and a good methodology than you can with any amount of prototyping.

For those who are not designers, visualizing the form and behavior of software that doesn't yet exist is difficult, if not impossible. Prototypes have been drafted into the role of a visualization tool for these businesspeople. Because a prototype is a rough model created with whatever prebuilt facilities are most readily available, prototypes are by nature filled with expedient compromises. But software that actually works—regardless of how badly—exerts a powerful pull on those who must pay for its development. A running—limping—prototype has an uncanny momentum out of proportion to its real value.

It is all too compelling for the manager to say, “Don't throw out the prototype. Let's use it as the foundation for the real product.” This decision can often lead to a situation in which the product never ships. The programmers are condemned to a role of perpetually resuscitating the program from life-threatening failures as it grows. Like the stack in which the first 25 bricks were placed haphazardly, no matter how precisely the bricks above them are placed, no matter how diligently the mason works, no matter how sticky and smooth the mortar, the force of gravity inevitably pulls it down somewhere around the 50th level of bricks.

The value of a prototype is in the education it gives you, not in the code itself. Developer sage Frederick Brooks says, “Plan to throw one away.” You will anyway, so you might as well do it under controlled circumstances.

In 1988, I sold a program called Ruby to Bill Gates. Ruby was a visual programming language that, when combined with Bill's QuickBasic product, became Visual Basic. What Gates saw was just a prototype, but it demonstrated some significant advances both in design and technology. (When he first saw it, he asked, “How did you do that?”) The Microsoft executive in charge of then-under-construction Windows 3.0, Russ Werner, was also assigned to Ruby. The subsequent deal we struck included having me write the actual program to completion. The first thing I did was to throw Ruby-the-prototype away and start over from scratch with nothing but the wisdom and experience. When Russ found out, he was astonished, angry, and indignant. He had never heard of such an outrageous thing, and was convinced that discarding the prototype would delay the product's release. It was a fait accompli, though, and despite Russ's fears we delivered the completed program on schedule. After Basic was grafted on, VB was one of Microsoft's more successful initial releases. In contrast, Windows 3.0 shipped more than a year late, and ever since it has been notoriously handicapped by its profuse quantities of vestigial prototype code.

In general, nontechnical managers erroneously value completed code—regardless of its robustness—much higher than design, or even the advice of those who wrote the code. A colleague, Clay Collier, who creates software for in-car navigation systems, told me this story about one system that he worked on for a large Japanese automotive electronics company. Clay developed—at his client's behest—a prototype of a consumer navigation system. As a good prototype should, it proved that the system would work, but beyond that the program barely functioned. One day the president of the Japanese electronics company came to the United States and wanted to see the program demonstrated. Clay's colleague—we'll call him Ralph—knew that he could not deny the Japanese president; he would have to put on a demo. So Ralph picked the president up at LAX in a car specially equipped with the prototype navigation system. Ralph knew that the prototype would give them directions to their offices in Los Angeles, but nothing else had been tested. To Ralph's chagrin, the president asked instead to go to a specific restaurant for lunch. Ralph was unfamiliar with the restaurant and wasn't at all confident that the prototype could get them there. He crossed his fingers and entered the restaurant's name, and to his surprise, the computer began to issue driving instructions: “Turn right on Lincoln,” “Move into the left lane,” and so on. Ralph dutifully followed as the president ruminated silently, but Ralph began to grow more uneasy as the instructions took them into increasingly unsavory parts of town. Ralph's anxiety peaked when he stopped the car on the computer's command and the passenger door was yanked open. To Ralph's eternal relief, the valet at the desired restaurant had opened it. A smile broke across the president's face.

However, the very success of this prototype demonstration backfired on Ralph. The president was so impressed by the system's functioning that he commanded that Ralph turn it into a product. Ralph protested that it was just a feasibility proof and not robust enough to use as the foundation for millions of consumer products. The president wouldn't hear of it. He had seen it work. Ralph did as he was told, and eight long years later his company finally shipped the first working version of the product. It was slow and buggy, and it fell short of newer, younger competitors. The New York Times called it “clearly inferior.”

The expertise and knowledge that Ralph and his team gained by building the prototype incorrectly was far more valuable than the code itself. The president misunderstood that and, by putting greater value on the code, made the entire company suffer for it.

If you define the boundaries of a development project only in terms of deadlines and feature lists, the product might be delivered on time, but it won't be desired. If, instead, you define the project in terms of quality and user satisfaction, you will get a product that users want, and it won't take any longer. There's an old Silicon Valley joke that asks, “How do you make a small fortune in software?” The answer, of course, is, “Start with a large fortune!” The hidden costs of even well-managed software-development projects are large enough to give Donald Trump pause. Yacht racing and drug habits are cheaper in the long run than writing software without the proper controls.

  • Creative Edge
  • Create BookmarkCreate Bookmark
  • Create Note or TagCreate Note or Tag
  • PrintPrint