Eben Moglen on NetApp v. Sun

The idea of Sun licensing its ZFS patents for GPLv2 implementation has been suggested, though.
More possible ZFS implementors could encourage more prior art researchers.

My favourite example is the powerboard/powerstrip. I heard that it was never patented - and
haven't we all benefited from it?

I understand the problem completely.


The problem is that although patents and copyrights are different laws entirely.. software is
covered by _both_.

What other endevour is realy covered by both copyrights and patents in the same way? 

Like I said: You can't copyright a engine block.. you can't copyright a windmill. (am I
right?) You can copyright a blueprint... but that isn't going to stop somebody from obtaining
your product and duplicating it for themselves. 

If you make a better mouse trap and don't patent it people can copy it and make as many as
they like. There is no copyright law that affects that. 



And it's like that for writing a book. You can't patent a storyline. Sure you can patent how
the paper is made or how the binding is done.. but that's not anything to do with the actual
writen words.



But with software it's covered by BOTH copyrights and patents. 

So you get double restrictions, both unrelated in law, but very related in their effect;
related strongly in reality.

So... patent law for software isn't going away any time soon. A theoretical compromise is that
authors/inventors can choose either or. 


So as part of the patent application they have to express their software patent in source code
form. If they can't produce their 'invention' in logical terms in software form then it
obviously is not ever going to apply to any real-world software and thus the concept is null
and void. 

And in the act of getting the patent the company is surrendering copyrights. Patent documents
(as I understand it) are in the public domain and thus you can use the contents in any way you
feel like. 

So in the act of getting the patent the software company is publishing their software into the
public domain. Nobody can use it without paying royalties, of course, for the duration of the
patent. But once that is up then that code is in the open and anybody can use it.


AND as another effect by having the source code aviable in the public domain this means you
can use this source code as proof in the court of law. If there is a patent dispute then you
can use scientific methods and expert testimonial to determine in a factual way weither or not
the source code used in a infringing product is violating the patent.

AND by doing this your going to establish some benchmark for infringement.. this means that
you should be able to create effective tools to help programmers analize code and avoid the
patents. Sure sure this isn't going to solve the problem inherent in software patents, but it
can help prove that your not willfully in violation and when patent trolls come into town you
can use these tools to help you decide which threats are credible and which ones are not. 

AND this would help curb the overabundance of software patents. Hundreds of thousands of new
software patents are made every year. If companies had to surrender large portions of their
source code to the public domain every time they tried to get a patent this would put a
definate chill on the whole thing. Thus only real breakthroughs will be patented. Only stuff
were the inventor is very sure that they can make more money by licensing then by selling.


Of course the life of a software patent should be shortenned from 14 years (or whatever it is
now) down to 7 years or so.

This will match about the lifespan of proprietary applications. By the time a company EOL's a
product then that's about the time any patents expire. 

For example the original Windows 2000 release would now be patent-free and projects like Wine
would be free to reverse engineer the hell out of it.


I hate software patents and understand their badness as much as anybody.  And I don't know of
any software patents that actually pushed software science further.


What I am interested in is maybe a compromise to make a broken system work in a way that,
while nobody is going to like it, would hurt much much less.

Yeah.  If Altavista or whatever had picked up on PageRank, then I doubt all the other neat
ideas Google has produced would have been produced, but that has as much to do with capital
investment and business cultures as it does the patent system.  I'm skeptical that this is a
general argument for the benefit of software patents.

IBM spending vast sums over many years to drag workable speech recognition into existence is a
better example, where the initial investment (I have heard) was justified by the potential
gains from owning the technology, just as the classic patent model suggests.

On the other hand, part of the reason that speech recognition was so expensive and took so
long is that they just didn't have the right tools and fast enough hardware.  Maybe without
patents, IBM wouldn't have spent so much on it, but then someone else would have spent much
less, much later to get the same results much faster... so that ultimately the arrival of
*usable* technology was only delayed by a few years.

It doesn't matter anyway, except as an intellectual exercise; even if software/math patents do
some good in some cases, it's entirely clear that they do more harm in more cases, and on the
net they impede the progress of science and the useful arts.

>Page rank patent? Isn't page rank more of a trade secret?

Umm, not at all.  It's very well studied, lots of academic papers, wikipedia has a detailed
description of the math: http://en.wikipedia.org/wiki/PageRank

It's also not exactly what Google uses these days -- their actual algorithms that have been
tweaked over the years to fight spammers are I believe trade secrets.  But PageRank is what
got them started, and whatever they use now probably still has the basic PageRank ideas in its
core.

> If not, could you please point me to a descrption of Google's core technologies that would
allow me to reimplement them myself?

I guess you can actually get pretty close if you read the PageRank papers, the GoogleFS paper,
the MapReduce paper, the BigTable paper, and some of what's available on their deployment and
sysadmin tools.  And have a few thousand geeks and infinite cashflow to implement it all and
apply it in clever ways to new problems.  The problem is that most of these tools aren't any
use unless you have exceptional implementations of them, and giant clusters to run them on.