[comp.society.futures] A software copyright solution?

n8445388@unicorn.WWU.EDU (Treon Verdery) (08/09/90)

fhapgood@world.std.com (Fred Hapgood) writes:

>A number of observers of the programming scene are of the opinion
>that the upcoming flood of software patents, taken together with
>the recent expansion by the courts of ownership rights in user
>interface copyrights, will cripple the development of the
>software industry in this country (and therefore the industry
>itself).

I may have thought of a partial solution to this problem, 
although I have no doubt others have thought of this idea as well.

If this is the first mention of this idea anywhere I formally declare
it to be in the public domain.

It is my understanding that exceptionally simple computer operations
like clearing the screen, or updating a variable within a loop have 
been "protected", and that they have been protected by describing
their exact process with a flow diagram. 

if one were to define these actions as objects, (as in object oriented
programming) and then train a neural net to approximate these objects,
could one assemble a program out of these neural approximations and avoid
specific violation of a software patent? "Yes, it updates cells in a 
spreadsheet.  No, I don't know specifically how it does it."

Also:

1) loading the neural net with something personal (say a representation
   of your name) on start up could make sure each group of weights
   within the neural net was unique to the user, hopefully _placing
   the responsibility for patent violation with the USER, not the
   software developer_, because the specific mechanism of the process
   would be user controlled.

   I would hope that this would be analogous to to an individual
   CHOOSING to put a cassette tape in a box, CHOOSING a radio station,
   and finally CHOOSING whether they were recording the Rolling Stones
   (tm) or the sounds of C-span (no tm).  I can imagine the whole
   thing being icon driven and taking 5 minutes the first time you
   use the software.

2) there are obvious disadvantages if a program executes neural
   net modules.  I suggest that a third party write a decompiler for the neural
   modules that the user could CHOOSE to install with the click of an icon.
   this need only be done once, and of course it would be up to the user
   as to whether it was automatically executed on startup.

This process seems cumbersome, but perhaps it will suggest a 
more streamlined alternative. I am not a "real programmer", so I need
instruction rather than flammage as to why this kind of thing wouldn't
work.  I'm not a "real lawyer" either, and so I'm interested in learning
the real legal basis for defeating the software copyright problem.

followups to comp.society.futures...feel free to distribute this article
in any way you like.

-Treon

treon@milton.u.washington.edu   OR  n8445388@unicorn.wwu.edu
 
 -one final note: I am interested in pursuing a master's in technology
  transfer.  If you have any advice, or can recommend any programs
  please contact me.