🚀 go-pugleaf

A couple of things to add:

1) on compilers "discovering" patentable ideas.

   Not quite, yet, but I believe that some people have demonstrated
that "finite differencing" applied to "stupid" string searches
generates more interesting forms of string-searching that have been
published as algorithms in their own right.  Not yet in today's
compilers, though, and not clear that "finite differencing" will ever
appear in any form more interesting that reduction-in-strength for
languages like C.

2) on "prior art".

   I think there's a difference, in terms of invalidating patents,
between "prior published art" and "prior unpublished art".  If Xerox
did it in 1960, but didn't bother to publish this information until
after AT&T filed for their patent, then I believe AT&T's claim is
still valid.  Of the examples of "prior art" published here, how many
of those were either published AFTER the patent filing date, or
learned under non-disclosure agreement?

3) on "repatenting the wheel".

  No, that is not allowed, but it may be possible to patent a "new
application" of the wheel.  (I don't make these rules, I just report
them.)  Thus, register allocation by graph coloring is a *new
application* of graph coloring.  These rules vary from country to
country -- for instance, much "smaller" inventions get patented in
Japan (but they're not fools -- they patent their "big" ones here,
too).

4) an obnoxious point of view.

  Is time better spent generating 5 new inventions per day, or reading
patents to learn how to solve our problems?  Arguments against/for this
obnoxious view include:

  a) patents are written clearly enough to get the patent granted, but
     no clearer.

  b) if the "cost" of re-inventing (as opposed to *copying*) the
     invention is substantially less than the "cost" of reading
     the patent and licensing it, is the system really more
     productive?  (i.e., the patent office maybe ought to hire some
     computer scientists.  But, many computer scientists also have
     degrees in EE, so there must be some overlap.)

  c) but -- there should be some reward for being first and being
     public about a useful discovery.

David Chase
Sun