dmr@alice.UucP (10/03/86)
People seem to have strange ideas about patents. A first, seemingly unknown, fact is that the legal content of a patent is in its claims ("What is claimed is:"). These are hard to read without a lawyer, but it is just not enough to say that Andy Bechtolsheim and Sun have a patent on not translating RAS addresses. It depends on the details. Second, most patents are (of course) about utterly trivial things that no one cares about, but even among those that have some value, most are not earthshakingly new or surprising or even, except in the legal sense, "non-obvious to those skilled in the art," particularly once they have been read. Third, patents are only rarely contested in court, and when they are, the patent frequently doesn't stand up. However the process is so expensive for everyone that it is worth avoiding if possible. There are newsworthy exceptions; the most spectacular recently was Polaroid's success in forcing Kodak out of the instant-photography market by winning a patent case. Fourth, in the electronics area, specific patents conferring exclusivity don't seem to be as important as they are elsewhere. For example, it is my impression that in the pharmaceutical industry, possession of a patent on a specific compound or process can be a very valuable thing. But in computers, it more usual either to arrange a license on reasonable terms or to get around the claims of a patent. There are exceptions here, too; DEC has been litigating for years with people about patents on their various busses. Especially for large companies, patents tend to be of value not for protection of specific ideas, but as a kind of commodity which can be traded with others in the form of cross-licensing agreements. If you have lots of patents, you are in a strong position to bargain once for all and not have to worry so much about costly lawsuits (either by you or against you). In all cases, what you do with a patent is an economic (game theory) decision. Consider what Sun might try. They could let things ride, since they're doing pretty well. They could try to get some license money from everyone who is using the idea, or they could offer cross-licensing agreements. Being aggressive might offer some real rewards (in a fantasy world, they might put their competitors out of business) but is likely to be costly. In the case of the SUID, Andy Tannenbaum is being overly dramatic. It wasn't patented in order to prevent Unix from being pirated, but instead because our lawyers are always asking us if we've had any bright ideas lately, and I disgorged this one. It certainly never entered anyone's mind that the value of the system as a whole rested on SUID. In fact, the existing consent decree required AT&T to license all its patents for reasonable fees; if someone had come out with a pirate Unix, the most that could be done with the patent (as opposed to trade-secret protection) would be to extract a little money. Early Unix licenses--I haven't read recent ones--mentioned that some patents had been applied for, and that the licencees might be responsible for fees. (The figure $50 comes to mind, but I could be wrong here.) Ultimately, only one patent issued, and AT&T decided that it was not worth while collecting any fee whatsoever. Effectively, anyone can now use the SUID idea for the price of a letter to AT&T in Greensboro to check on what formalities might still be required. Incidentally, contrary to what Andy says, and also in spite of Benson-Tabbot (the Supreme Court case that Bell Labs lost), our lawyers have been telling us for many, many years that software is indeed patentable. Ken Thompson's patent "Regular Expression Search Algorithm" must have been filed around 1968, and it is full of 7090 assembly language. (On the other hand, its diagrams are full of AND gates, as indeed are the diagrams of the SUID patent). Dennis Ritchie --