Lapas attēli
PDF
ePub

OK; now, the other thing that has come up a number of times and came up this morning is what we call reverse engineering, which has been a concept understood in the industry which has always been considered perfectly fair. This is the equivalent to individual rights of a copyrighted biography. You can get a copyright in a biography.

A second writer can still write a biography about the same person; he just can't use the same words. He has got to express it in a different manner. And so with chip, we want to take the same concept and just apply the normal fair use doctrine to fair reverse engineering.

And the way you do that is you don't copy the thing directly for a mere $100,000; what you do is you study the operation of the chip. You take that chip and you say, how does it work? We publish what the specifications are, and the guy says, OK, I can do that better than they did it.

And so you can implement those exact same functions. You just can't do it with that specific picture. You'd use a different picture. And that's perfectly fair in our minds, and I think it's perfectly fair under the normal copyright doctrines. By doing that, you can reduce the cost. You can improve the performance. You have seen how one person did it. The second time around, you're going to do it better.

Now, the difference between reverse engineering and direct copying is that reverse engineering is going to cost about 25 percent of the original design, and it's also going to advance the state of the art. You're going to have a better chip; whereas the copying, you're just directly copy. So you're really not advancing the state of the art. All you're going to do is copy it, manufacture it cheaper, and therefore discourage the design of innovative chips.

The last thing I want to go to is the economics of this thing. The chip is not very useful by itself. You have to have a family of these chips, and you have to have additional products to help the customer design a system.

So the typical cost of a chip family like this one, which happens to be what is known as the 8086, will cost in the neighborhood of $80 million, so that that is broken down to about $4 million, what I call the main chip, which is this CPU, central processing unit. So that was $4 million by itself, but it is not useful unless you have done a study of the market.

So how can a broad group of customers use that so you can sell many of these chips? That is another $36 million. Then you are going to have to develop additional chips, and then you are going to have to develop development tools, which is a hardware system, a computer system, to allow your customer to use that chip, and the software that goes with it.

So this is a major development project which costs about $80 million. It will cost about $10 million, just ongoing maintenance, to solve the customer's problems and upgrade the thing.

Mr. KASTENMEIER. Mr. Dunlap, can you be more consistent than that? Can you give us a case of another company that you sell this to and what this chip will do with the finished electronic unit so that we can actually see how it is integrated into commercial purposes?

Mr. DUNLAP. The most common example is we have a slight variation of the chip which is shown as the 8088. That chip is used by IBM as the brains of its personal computer, its latest personal computer which has just come out in the last year or so. The brains behind that is an Intel chip.

Mr. FRANK. Are there ever times when it makes sense for a company to reverse its own chip?

Mr. DUNLAP. In fact, we do that. We make it smaller. We add functions, and to some extent-the case of the IBM personal computer is an example when we first came out with this 8086 which takes 16 bits of information in at a time and then processes. It gives 16 bits of information out at a time. IBM in the personal computer and a number of other people said, but your last chip had 8 bits, and I did a lot of work with 8 bits, I have a lot of software. So I like the idea that you process 16 bits internally, but I want to give you 8 bits in-convert it to 16 bits for internal use, and give me 8 bits back. You are converting it and using reverse engineering to modify it.

Mr. SAWYER. Would it be fair to say, so that I understand what you are saying, that reverse engineering would be designing around a patent? In other words, we know a light switch turns on and off a light and that is a function, but we could figure out another way to do it mechanically?

Mr. DUNLAP. That is exactly the same.

So reverse engineering. This chips costs about $1 million, takes about 12 years and improves the state of the art. A direct copy costs about $100,000, takes about 3 to 5 months, and doesn't really advance the state of the art. It is really the same thing.

Now either of these people-they don't have any of this. All they have to worry about is copying this chip, and it is much smaller, but we have developed the market for them and we have developed all the tools. So the customer buys the tools from us, and he buys the cheapest product.

And so, the net effect of this is that if we continue to allow people to copy our more advanced circuits—and there's chips more advanced than this, certainly in design, some in production-what we have to do is we have to say, OK, it is going to cost me $80 million, it is going to cost me $10 million to continue to update. Now what is going to be my return? What are my average selling prices going to be? What are my margins going to be?

And today when we do this, we look out and we say, it is going to take 3 to 5 months to copy it. At that point they are going to substantially reduce the price because they haven't had to invest any of this. They don't have to recover any of these other costs. Therefore, is it even worth doing?

And part of our decision has to be based on a copy coming out of the market. Maybe we shouldn't do the chip. So the ultimate result is we discourage innovation, and maybe the personal computer will not advance because we can't advance the chips.

Mr. FRANK. Is there a gray area between reverse engineering and copying? Would you have to subpena any product?

Suppose somebody made a copy and made just enough changes to new product. Would you then have to subpena them and require them to show a work product?

Mr. DUNLAP. I think there is always going to be a gray area. The majority of cases are going to be straightforward.

Here we have this picture of 100,000 transistors. Is someone going to lay out 100,000 transistors and have them look exactly the same as this? Highly unlikely.

Mr. FRANK. I agree, but the question is can you do a copy and make just enough changes so that you might make it look like you were reverse engineering, but you really weren't?

Mr. DUNLAP. Well, that is going to be the problem. The guy who copies it tries to get around it, and that is what you have to-Mr. FRANK. Would you then require that the work product be there so that they have to show how they manage to arrive at that, how they get that end product?

Mr. DUNLAP. That is what you have to do.

Mr. FRANK. Do we have to address that? Would that be automatic that the counsel for-

Mr. DUNLAP. Mr. Frank, you have hit it exactly on the head. When there is a legitimate job of reverse engineering, there is a very big paper trail, there's computer simulations, there's all kind of time records, people who have spent an enormous amount of time understanding and figuring out how to make the design.

Mr. FRANK. It is not the extent of the change, but the extent to which the work can be documented and the corrections can be documented.

Mr. DUNLAP. Correct, whenever there is a reversing engineering job, there is a very big paper trail that cannot readily be fabricated.

Mr. FRANK. It would be someone who would do what you do and someone was accused of a copy and the defense wouldn't know if they reverse engineered, then they would have to have the burden of showing if they couldn't produce the paper trail?

Are you against that?

Mr. DUNLAP. The point you have raised is specifically addressed to the other body by Mr. Vadasz, who is an officer of Intel, and with your permission I would like to make that letter that he sent to the other body a part of this hearing also.

Mr. FRANK. I would ask that this be done. [The letter follows:]

INTEL CORP.,

Santa Clara, Calif., June 23, 1983.

Hon. CHARLES MCC. MATHAIS, Jr.,
U.S. Senate,

Washington, D.C.

DEAR SENATOR MATHIAS: I understand that two questions have been raised, concerning S. 1201, on which I would like my comments to be made part of the record. I am an electrical engineer by training and have spent the last twenty-two years in the solid state electronics area involved in the design and development of semiconductors. As a result of my work, I was made a Fellow (the highest technical position) of IEEE. I received this honor for leadership in the design and development of semiconductor memories and microprocessors. I feel that gives me the authority to speak out on these issues.

First, it has been suggested that a copyright on a set of masks can somehow monopolize electronic circuitry so that later manufacturers will be prevented from using essential designs. In the same view, it has been alleged that any engineer with the requisite skill, working on a given circuit, will tend to converge on a single most reasonable mask layout.

This is completely contrary to the experience of engineers in the semiconductor chip industry. For any desired function, there will always be a large number of different good layouts. A copyright on one layout will not keep engineers from using other functionally equivalent but visually dissimilar layouts. Engineers do not converge on a single most reasonable layouts because no such thing exists. When an engineer creates his own layout instead of copying someone else's, he invariably comes up with something that looks different-probably even to a casual lay observer, but certainly to a trained eye. The likelihood of two engineers coming up with the same chip layout is equivalent to the likelihood of two college students independently writing the same essay on a final exam.

Second, it has been said that even very subtle mask changes may represent significantly different and original designs. This is true. It has been further said that exactly the sort of tests that demonstrate such differences are specifically disallowed as defenses in copyright infringement cases. I do not believe this is true, for I have been informed otherwise. But I feel that evidence of this type should be allowed in semiconductor chip copyright infringement cases and hope that the legislative history of S. 1201 would include a statement endorsing use of expert testimony to show subtle functional differences in circuit layouts.

Finally, a point deserves mention that has a bearing on both of the foregoing points. When a company decides to become a second source for a chip already on the market, it will probably want it to be equivalent to the first chip not only functionally but in terms of specifications and test data; that is, the second chip would be so fungible with the first chip from a production standpoint that it would not make any difference which one was placed into the equipment for which the chip is targeted. In these circumstances, a chip designer may feel that the fewer design or layout changes that are made from the first chip, the less likelihood there will be of a nonequivalence in specifications. This would lead to similarities in layout and appearance, but even when this happens, it is reasonably easy to tell the difference between a slavish copy and a reverse engineering job. Whenever there is a true case of reverse engineering, the second firm will have prepared a great deal of paperlogic and circuit diagrams, trial layouts, computer simulations of the chip, and the like; it will also have invested thousands of hours of work. All of these can be documented by reference to the firm's ordinary business records. A pirate has no such papers, for the pirate does none of this work. Therefore, whether there has been a true reverse engineering job or just a job of copying can be shown by looking at the defendant's records. The paper trail of a chip tells a discerning observer whether the chip is a copy or embodies the effort of reverse engineering. I would hope that a court deciding a lawsuit for copyright infringement under this Act would consider evidence of this type as it is extremely probative of whether the defendant's intent is to copy or to reverse engineer.

Sincerely yours,

LESLIE L. VADASZ,
Senior Vice President.

Mr. DUNLAP. I think that the short answer in terms of the majority of the cases of copying, it is going to be something like obscenity. You will know when you see it.

Mr. FRANK. That would make me very nervous. I like the other— I like the long answer better.

It seems to me there has got to be some way we can avoid that in the paper trail. It seems to me that that ought to be a factor. Otherwise, it becomes too subjective.

So it is not just the nature of the changes, that somebody can maybe counterfeit those or make just a few cosmetic changes and try and have it finished. It is a fact that he would have to showyou would have to show how it is worked out. That would make the difference. I don't think there is much of a paper trail, but I am not an expert.

Mr. DUNLAP. The fabrication of a paper trail like that is almost impossible.

Mr. FRANK. Well, if you do that, you might as well have gone and done the original work?

Mr. DUNLAP. That is correct.

Mr. KASTENMEIER. Presently or in the future, under such a bill could there be an obligation imposed on the part of the original designer to incorporate one or more unique codes that would identify that design as against replication?

Mr. DUNLAP. Well, that is interesting. These chips are so complex that usually the engineers do that for us by accident. They don't try to.

This particular chip was copied. Our design engineers made some mistakes the first time around, and when they fixed the mistakes they fixed them in quick manner. In other words, they put in-in this memory but they had to change some bits. They didn't need some of them. So what they did was they just cut them off so they wouldn't be used. It was a quick way to do it.

The person that copied it copied all these unused things. So they copied our mistakes, which is quite often the case, where they will copy a circuit that has no function.

Mr. KASTENMEIER. That would be a distinction between an original design and replication?

Mr. DUNLAP. That is correct.

Mr. STERN. Mr. Chairman, if I may, the story of that incident is described in a Washington Post story, which you may have read here a couple of months ago. If it would suit the chairman's convenience, we would be glad to make a copy of that story from the Washington Post about copying the mistakes a part of the record in this proceeding.

Mr. KASTENMEIER. Without objection, we will receive that and make it available to the members of the committee.

[The information follows:]

[From the Washington Post, May 2, 1983]

BATTLING TO INNOVATE AND EMULATE: INTEL VERSUS NIPPON ELECTRIC

(By Dan Morgan)

Peering into a microscope at a greatly magnified computer chip one day last August, Peter Stoll of Intel Corp. saw something startlingly familiar. In one of the tiny cells, two transistors were disconnected from the rest of the chip, and dangled uselessly in their bed of silicon.

Stoll, 33, a chip designer, recognized the defect as a small last-minute repair job he had performed on Intel's 8086 microprocessor several years earlier. It had worked, correcting the minor flaw in the chip's logic, and the 8086 went on to become phenomenally successful as the "brain" in a wide range of business computers, robots and industrial machinery.

But what startled Stoll was that the chip under the microscope was not Intel's. It was a product of Nippon Electric Co. (NEC) of Tokyo. Stoll concluded that he was looking at a Japanese copy so perfect that it even repeated the small imperfection in the original chip.

Intrigue of that kind in the $13 billion-a-year global market for computer chips has led to U.S. accusations of unfair Japanese practices, ranging from copying to protectionism. Critics of Japan say that its efforts to gain supremacy in computer chips, perhaps the single most important technology of the Information Age, are typical of the methods employed by "Japan Inc."

"We're at war, no doubt about it," said a computer scientist from a large U.S. research laboratory. "If I had money in 'Silicon Valley,' I'd get it out. . . . It's just like any other war zone."

U.S. politicians are in a mood to strike back.

Democratic Reps. Don Edwards and Norman Y. Mineta, from California's socalled Silicon Valley area, have introduced a bill to give copyright protection to chip designs. They say the measure is needed to stop "pirate firms" from "flooding markets with copied designs that undersell the innovating firms."

« iepriekšējāTurpināt »