Re: Compcomm: Draft "IUCr Computing Commission's opinion on

[Vincent Favre-Nicolin <vincefn@users.sourceforge.net>]

	Hi (I am now receiving the list emails)

   A few points to clarify.

1)  the original aim of patents was not to protect the right of 
inventors, but to ensure that inventions would not be kept secret. 
Giving legal protection to inventors is just a way of achieving that 
which was introduced IIRC in the 19th century. Eventually, the only 
thing that matters is the spread of technical & scientific knowledge.

   This directly raises the problem of "trivial" patents (such as using 
exclusive OR when a cursor overlays a character, etc). Obviously such 
small things should not be patentable. Practically (see the patent 
horror gallery), the Patent Offices are not technically competent to 
evaluate if it is trivial or not.
   And this is a problem specifically with software because related 
patents tend to be trivial.

2)
Simon:
> Someone can invent an algorithm, just as they can invent a new kind of
> vacuum cleaner or an anticancer drug. Why should this not be subject
> to protection?

   There is a very large difference between those examples: how long is 
it going to devise a new drug ? 5 to 10 years, including all research, 
testing, validation by drug administration, implying many people. Same 
thing for many industrial production processes, which require a lot of 
"man-years" of work to obtain the result.
   On the other hand, how long does it take for a software part ? Much, 
much less. In a large number of cases (the trivial patents) the idea 
that is patented can require a few hours (and the patent is accepted). 
There _are_ still complex algorithms which require more work, but it is 
almost systematically 10 or 100 times less "man-years" of work than 
industrial-grade patents.

   So should all these methods also be given a ~20 years patent ?

   Bottom line (IMHO) is:
   - there exists no provisions against trivial patents
   - software patents (even non-trivial) require much less time to 
devise (remember - only the principle of the algorithm is the subject 
of the patent, not the software itself, which could take much more time 
to implement/test...) than industrial methods
   - the software industry has been flourishing in the last 10 to 20 
years, without these patents in Europe - so why do we need these ?
   - algorithm knowledge (know-how) has also spread during this period - 
so knowledge spread (the original aim of patents) would not benefit 
from software patents
   - wide software/algorithms patentability would seriously reduce the 
ability of independent software developpers or small corporations or 
scientists to develop new software (no money to defend oneself in case 
of "trivial" or "prior art" patents, need to "buy" the right to work on 
a less-than-20-years-old scientific algorithm).


  More generally, it seems to me in the spirit of the original patent 
laws that only complex processes, which require a long development and 
testing, should be patentable (since the duration of the patent is ~20 
years).
  Yes, some very elaborate algorithms may qualify for the above, but 
without any provision on a required complexity for each patent in the 
law, it seems to me that the loss of legal protection on these (a 
handful of algorithms?) is way, way less important than the drawbacks.


   What would crystallographically be if experimenting on direct methods 
/ monte-Carlo / fourier recycling was patented for 20 years? That would 
slow research. 

	Vincent
-- 
Vincent Favre-Nicolin
Université Joseph Fourier - Grenoble, France
http://v.favrenicolin.free.fr
ObjCryst & Fox : http://objcryst.sourceforge.net