slf@cs.mu.OZ.AU (Stewart Forster) (06/26/91)
Hi, Having doodled with compression algorithms for the last two years, I have come up with a few ideas which I thought were original, and at least one still is (to my knowledge). Now I'm no expert on legal matters, but quite often I seem to be hitting this brick wall of people saying "But that's already been patented". What I want to know is what does this slug called "patent" cover. Does it cover the idea or the implementation or what?? Does it mean that everything from the concept of looking up strings in a dictionary, or using arithmetic coding or huffman, or even all the way back to the fundamentals of data compression itself. If this is so, we all might not be able to use data compression schemes because idea of compressing data was first thought of Shannon (or whoever it was exactly). Let me take this idea a step further. What if someone had patented the idea of putting a motor on a four wheeled carriage. If this guy had his way, we would all still be driving his "idea" of what a "car" should be. This is because, the patent is too restrictive and does not allow room for advancement. The above example hasn't occurred, but it seems to be happening with data compression schemes. eg. "I own the patent for looking up duplicate strings ahead of the current position, and placing them into a dictionary". This sort of broad based generalization can cover 1/4 of the possible data compression schemes. Why not take it one step further, eg. "I hold the patent for the idea of using duplicated strings as a tool for data compression". This can cover every Lempel-Ziv style data compressor, and every Markov model higher than 0th order. So, if the the two above examples are ridiculously general, and we specify some more, we are about to descend into the realm of implementation specifics. eg " I own the patent of using hashing into a circular buffer to find duplicated strings from the previous text, and then outputting the length and position of that duplicated string". This has two flaws which make the patent either restricive or useless. It is restrictive, because if I then decided to use 0-order markov coding of the lengths and/or positions in the circular buffer, am I breaching the patent stated above, because I have improved compression with this new method, and while the fundamental idea is still the same, the implementation to get to the coded output is more advanced. This relates to the car analogy 2 paragraphs above. Alternatively, the patent above is useless, because, what if I then decided to implement looking up into the previous text, not as a circular buffer with hashing, but as a modified Patricia trie, has Fiala and Greene have done. Obviously the patent above doesn't cover this eventualaity, and if it were generalized to do so, then we are back to the car analogy again. I am tired of trying to dodge peoples toes, and either I do some stamping, or the very good algorithm(s) which I have already implemented will go to waste just because a portion (or even 90% of them) is similar to what other people have done (idea-wise, not necessarily implementation), but that other extra bit is what provides the compression-performance/speed edge over the original idea. Has Phil Katz patented his PKZip algorithm? If he has, then we should all have been able to see it by now, or is he scared of another patent scam again, and that's why he is keeping the code so secret. The same probably goes for Robert Jung, with ARJ 2.00. Am I expecting too much, being a Unix user, and used to having freely distributable source code to modify and redistribute as long as I aknowledge the original author of the code. This sort of environment promotes advancement of ideas as these ideas come collectively from the everyone in the world, rather than just one solitary patented mind. Will someone please lay my battered soul to rest, it has been hitting too many brick walls lately, some of which I pray are imaginary. Should I just remove my diligence from the world of compression, deplete the overall effort going towards producing the best compression algorithm possible. Or should I just pound along and hope that anything I do come up with will hold it's own in a court of law, when I'm no lawyer. Patents are an ugly slug, especially when they inhibit experimental research. Stewart Forster. -- Stewart Forster (slf@cs.mu.OZ.AU) Any opinions are strictly my own, as if they weren't!
brnstnd@kramden.acf.nyu.edu (Dan Bernstein) (06/26/91)
You should check any good textbook or introductory article on patents. In article <9117710.2823@mulga.cs.mu.OZ.AU> slf@cs.mu.OZ.AU (Stewart Forster) writes: > Let me take this idea a step further. What if someone had patented > the idea of putting a motor on a four wheeled carriage. If this guy had his > way, we would all still be driving his "idea" of what a "car" should be. Patents only last 17 years. > Patents are an ugly slug, especially when they inhibit experimental research. If that's all you're worried about, stop worrying. Patents never apply to use ``for the sole purpose of satisfying philosophical taste or curiosity, or for instruction and amusement.'' If you're just playing with a patented invention in your back yard trying to make it work better, you're safe. Just don't try to sell it or give it to other people to use. ---Dan