Richard Stallman on Software Idea Patents (Groklaw)

Posted Oct 18, 2004 21:04 UTC (Mon) by lakeland (subscriber, #1157) [Link]

The definition of what is patentable is not related to being trivial, but
to what is obvious. I seem to recall that hairclips were patented at one
stage, and yet they are 'trivial' to produce. They were simply an
extremely good idea. Something is patentable if it is nonobvious, not if
it is nontrivial.

The classic argument for software patents when confronted with "But that's
the obvious way of doing it" is "Yes, it is the obvious way of doing it
_now_, but you didn't do it before now, did you?" To me, this presents a
similar but different test. If a skilled programmer, presented with the
same problem, comes up with a solution that is substantially similar then
the patent is invalid. Because lawyers would argue over 'substantially
similar', I'd define it to: Any ideas expressed in the skilled
programmer's implementation must be struck out of the patent. It would
probably be worthwhile employing several skilled programmers per patent,
and I think that three days sounds reasonable, one to research, one to
implement, and one to examine similarities. That works out to nine
chargable days, or very roughly $5000 per patent appliciation.

You know, being given lots of little "implement this idea" tasks would be
a not bad job.

Posted Oct 23, 2004 20:39 UTC (Sat) by giraffedata (subscriber, #1954) [Link]

Patents don't have and aren't meant to have anything to do with how hard it is to implement something. They're about how hard it is to invent something.

If you were talking about copyright, it would make sense to have a standard of "how hard is it to write the copied bit from scratch?" But not for patents.

The cost of inventing most software inventions is truly trivial. You get an idea and then it's quickly obvious that it solves the problem. The invention is done before you ever write any code. That's why it's unreasonable to give them patent protection.

In the early days of patents, you usually could not reasonably claim to have invented something until you had built at least one working prototype. And in today's world, you can't claim success in some inventions (e.g. drugs) until you've done millions of dollars of testing. But most software inventions are rather different.

One of the greatest inventors of all time, Thomas Edison, has a famous saying, "invention is 1% inspiration and 99% perspiration." It's the perspiration that the patent recognizes. But most software inventions are 99% inspiration and 1% perspiration.

Posted Oct 18, 2004 19:59 UTC (Mon) by bjlucier (subscriber, #5281) [Link]

Re:

For the record the best way to compute n! is as follow:
1] write the list of integer from 1 to n:
2] multiply the first item of the list with the last,
the second with the last but one, etc.
3] Restart 2] with the current list until there is an only
element left.
4] the remaining element is the factorial.

This assumes you use asymptotically fast multiplication at
each stage.
For example for 8!
1 2 3 4 5 6 7 8
8 14 18 20
160 252
40320

This is getting way off topic, but it's faster to compute factorials using the method from Section 6.5 "How to compute n!" of Fast Algorithms by A. Schoenhage, A. F. W. Grotefeld, and E. Vetter. See also P. B. Borwein, "On the complexity of calculating factorials," J. of Algorithms 4 (1985), 376-380.

Posted Oct 18, 2004 21:22 UTC (Mon) by maney (subscriber, #12630) [Link]

If, instead of patenting LZW, Unisys had copyrighted the pseudocode and claimed that any implementation was a derived work, the technique would essentially never be useable.

Not at all. Your copyright mantra is abstraction, filtration, comparison. Copyright covers expression of an idea, not the idea itself. I think you're having the same misunderstanding as poor SCO in their doomed quest for discovery of every single change that ever occurred in AIX and Dynix. Derived doesn't mean all the same things in copyright as it does in our everyday usage. At least that's how I understand it after a year or more of informal education in the course of following the SCOundrels ill-fated attempt to cash in their chips by annoying IBM. There aren't so many chips left these days...

Anyway, what makes you think Unisys (or someone) hasn't got a copyright for one or more pseudo-code versions of LZW?

Posted Oct 18, 2004 21:32 UTC (Mon) by vmole (guest, #111) [Link]

A specific implementation of an idea expressed in psuedo-code idea is no more a "derived work" than _Macbeth_ is a derived work of the text "So, there's this prince in Denmark, and his uncle kills his father and marries his mother, and he gets all angsty about it."

As far as why people object to software patents, I think you're wrong. Most people probably don't object to the LZW patent (although they might object to the way Unisys submarined it), or the RSA patent, because there is actual creativity there. What people object to is that patents are being granted for the software equivalent of "a crescent wrench": tools that are in common use in the field, but which the patent examiners are not sufficiently knowledgeable or trained to notice.

Posted Oct 21, 2004 19:45 UTC (Thu) by pivot (guest, #588) [Link]

Yes SHOULD is correct. I didn't know ideas were patentable in the US.