NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/misc/intprop/860

Internet Message Format | 1992-08-18 | 3.5 KB

Path: sparky!uunet!wupost!sdd.hp.com!think.com!barmar
From: [e-mail] (Barry Margolin)
Newsgroups: misc.int-property
Subject: Re: Copyright and Derivative Works
Date: 19 Aug 1992 00:22:07 GMT
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 65
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <[e-mail]>
NNTP-Posting-Host: telecaster.think.com

In article <[e-mail]> [e-mail] (G.D.Shaw) writes:
>   My point is that there are cases where a text can
>   (a) be derived from another, (b) be useful to the recipient but
>   (c) which cannot be shown to be a derivative (or even be similar)
>   to any standard of proof whatsoever.

There are lots of things that are true but can't be proved.

>   What I was challenging was that "text A is derived from text B"
>   implies that "there will be discernable similarity between text
>   A and text B".

You're being pretty vague.  I presume you're talking about transformations
such as encryption.  Simply looking at the two texts will not make their
similarity apparent.  However, the court may be able to force the defendant
to decrypt the text (much as it can force him to open a safe, or reveal
other pertinent information), and then the similarity should be apparent.

A less clear case is compilation.  In this case, there's no simple inverse
transformation, but an expert witness could analyze the source and binary
and be able to determine whether the binary is likely to be the result of
compiling the source.

>The fundamental problem is this: the information content of a text
>is a relative, not an absolute quality.  It is possible for a text
>to be derived from an original without carrying any information
>about the original.

Could you give a useful example?  I can think of some pretty silly ones (in
which case they don't satisfy your criterion (b) above).  For instance, a
transformation where all words are replaced by randomly chosen words of
equal length.

I suppose that a series of useful transformation could have the property
you suggest (this reminds me of the word puzzles where you must transform a
word into another word by changing one letter at a time, each step
resulting in a valid word).

>              It is also possible for subsequent text that
>was not in any way derived from or contaminated by the original
>to convey enough information to reconstruct the original in its
>entirity.

And it's also possible for a random text generator to recreate the complete
works of Knuth (the modern version of the infinite monkeys).  In both
cases, it's unlikely enough that it can usually be ignored.

>Finally, perhaps I should make clear that I am not talking about
>'practical' law here - I am talking about what the judges should
>uphold, not what they would.  I am also aware that few (if any)
>judges are in the least bit qualified to judge what is right or
>wrong in this area.  (Nor, for that matter, are most of our
>legislators.)

If they can't base their judgement on direct comparison of the texts, then
they will have to resort to other types of evidence.  There are many cases
where direct evidence is not always available or reliable.  For instance,
an eyewitness who sees the defendant run the transformation program, or
hears him talk about it in the hallway, can be more useful than direct
comparison.
--
Barry Margolin
System Manager, Thinking Machines Corp.

[e-mail]          {uunet,harvard}!think!barmar