Viral permissiveness

[Donovan Hawkins]

On Wed, 4 Feb 2009, Steve Thomas wrote:

> Under this interpretation of copyright law, isn't some "provenance"
> data structure or markup associated with a source file required such
> that, given a specific character of the file (highlighted by a cursor
> in a text editor, say), it is possible to ascertain information
> including:
>  1) the license pertaining to the Work that character came from
>  2) the owner of the copyright in that Work, if there is one

>
> In simple and infrequently changing cases, this could perhaps be
> maintained manually. Otherwise, text editors wou
> ld be required that kept track of such metadata for their subject
> files and for their buffers.

But in this case the whole is greater than the sum of its parts. A 
person can have copyright on 1000 characters of source code, but they 
don't have copyright on each individual letter taken separately because a 
single letter is not original.

As the text gets sliced and diced, the "spirit" of copyright is still 
there and still taints the derivative work but may not exist in any 
specific letter. For example, extensive global search-and-replace combined 
with code obfuscation tricks might eliminate every single letter of the 
original but you would still be subject to copyright and the license 
restrictions of the original. No judge would permit alterations which can 
be 100% automated to create an original work. If they did, the compiled 
executable would qualify (as would translations of literary works).

On the other hand, a careful rewrite which reimplements the algorithm "in 
your own words" can escape copyright and produce a new original. The 
extreme example of this is setting up a Chinese wall where one person or 
team converts the source code into an algorithmic description which is not 
protected by copyright (but might be protected by patents) and a different 
team who has never seen the source code reimplements the algorithm.

There's a fine line between mechanical modification and real rewriting 
that a text editor couldn't possibly understand, nor could it monitor the 
buffer in the typist's brain which allows him to delete some text and 
retype it again with trivial modifications in another location.

Fortunately, this seems like more of a philosophical problem than an 
actual one. Humans can pretty easily compare both works and decide whether 
one incorporates the other in a meaningful sense. Obfuscation and text 
replacement can be pointed out to a judge by the original author, and he 
can render a reasonable decision that most of us could predict given the 
facts of the case. The only ambiguity lies in our inherent inablity to 
precisely define how many changes are required to completely obliterate 
the original work, making the answer somewhat subjective in difficult 
cases. Reverse engineers use Chinese walls precisely to avoid this 
question.

In short, this is a question that human beings can generally answer with 
ease but which would be almost impossible for a computer to answer in many 
important cases. The concept of "copying" involves more than the act of 
transcription; it is really about whether the text of the original drove 
the execution of the copy and acted as the primary creative element for 
the resulting text. Word-by-word or even line-by-line rewording is still 
plagiarism to most people because it is a "mechanical" process even when a 
human being performs it, yet a machine would be hard pressed to recognize 
it as copying.

As for the case of a person taking part of a GPL'd work and treating it as 
being under BSDL because that portion originated under BSDL, the judge is 
probably not going to object if the text is immediately identifiable as 
being part of the original BSDL work with only trivial modifications. My 
impression is that even the people who believe relicensing is 
techincally legal do not believe you would have standing to sue for 
copyright infringement when there is no original work added...the 
arguments I've read online have been largely philosophical ones as I 
understand them.

And if the original BSDL work has been sliced and diced to the point of 
identifying where individual letters came from, then the slicing and 
dicing itself probably deserves copyright protection.

---------------------------------------------------------------------------
Donovan Hawkins, PhD                 "The study of physics will always be
Software Engineer                     safer than biology, for while the
hawkins at cephira.com                   hazards of physics drop off as 1/r^2,
http://www.cephira.com                biological ones grow exponentially."
---------------------------------------------------------------------------