Subject: Re: termless copyright and patents
From: Thomas Lord <lord@emf.net>
Date: Fri, 29 Sep 2006 09:35:38 -0700
Fri, 29 Sep 2006 09:35:38 -0700
simo wrote:
> One thing is sure anyway. A Patent is often built on many claims, you
> are given a covenant only for the claims that apply to the software as
> distributed. So if you, by deriving the software start infringing in
> other claims, you are not covered by the covenant.
>   

The GPL takes all of Mr. Rosen's fun out of leaving the copyright
business and going to the patent business because it requires patent
holders to make a covenant expressed in terms of copyright permissions.

Would GPL allow me to cut an XML lexer out of a web server and
inject it into a database engine?    Clearly that is the intent.   At what
point are related patent claims captured by the GPL covenant and at
what point not?


>> Sure, they are better off using the GPLed code first and, often, that
>> will not
>> mean that they have to GPL all of their own code if they don't want
>> to.
>>     
>
> As you know this one really depends on the patent and the code, so you
> can;t really generalize.
>
>   

I think you can generalize enough to say this is "very often" the case
and to observe that large-portfolio patent holders who distribute
large quantities of GPLed code will have a very hard time identifying
when they are invoking this consequence.




>> But supposing they haven't been using GPL, and that problem is trivial
>> to
>> correct, if the patent holder has been damaged *at all* it isn't by
>> much.
>>     
>
> How so? Damages are often calculated based on the amount of money made
> by the infringer by infringing the patent. It does not matter if the
> same patent was available for free for other uses, the patent holder
> decides how to license and make it available, if you loose in defending
> your right to use it then damages can well be asked.
>
>   

Google for the "Georgia-Pacific factors" and look for one that talks 
about things
like "causation" and "but-for".




>>>> Holders of defensive portfolios, who distribute lots of GPLed
>>>> code, are at interesting risk here: their defense stance is greatly
>>>> weakened relative to other firms not distributing lots of GPLed
>>>> code.
>>>>     
>>>>         
>>> No, that's wrong again, the patents will not be available to
>>> competitors, unless they use the patent holder GPLed program.
>>>   
>>>       
>> Which is not much of a restriction.
>>     
>
> Oh well you can't say that, go ask MS to release the source code of
> their OS under the GPL because they infringe on a patent about managing
> the Virtual Memory ...
>
>   

No, I mean it the other way.    MS can very often cure the infringement
by releasing much, much less than all of their OS.



>> Right, they must go through the fairly trivial exercise of deriving
>> the code
>> which embodies the patent from sources which the patent holder has
>> conveyed.
>>     
>
> You assume everything is trivial, I wish it was so, we would have
> fantastic software today :-)
>
>   

?

That sort of thing happens all the time.   It's one of the big advantages of
having lots and lots of GPL software.   In my experience (not for patents
per se but just to extract and radically adapt complex components) it's one
of the fastest ways to get code written.




>> And if they have previously failed to do that in some instance, but
>> the 
>> problem is trivial to correct, the defensive patent isn't worth much.
>>     
>
> non sequitur, the infringment was done by non-GPLed and non-derived
> software, remedies to past infringement are separate from license for
> future distribution.
>
>   

Right.   See "Georgia-Pacific".   What *would* have happened if the 
infringer
was aware of the issue at the start?   How much extra money did they make
by not using the GPL code in the first place?   How much did the 
patentee lose?



>> It's interesting that you should bring this up because it applies to
>> purely 
>> GPLed programs as well.    Suppose that Acme Co. conveys a Linux
>> kernel
>> embodying one of their patents.    I, on the other hand, do a fresh
>> GPL
>> implementation of their invention for use in my lisp machine kernel.
>> If Acme Co. has not themselves conveyed my lisp machine kernel, 
>> then neither may I.
>>     
>
> Correct, you will not be covered. What's strange in it?
>
>   

I would say sad, not strange.    This is one reason to like the trend
towards covenants that are broader than the GPL requires.

It's sad because it is against the spirit of software freedoms.   I can
not study one free program, learn from it, and write a new free
program based on what I've learned without risking patent infringement
even though the patents in question are already embodied in the
program I studied.

Are you really going to be happy with a situation where, say,
GNU Awk has a very fast string library but you are prohibited from
using those same algorithms in GNU Classpath?


>> Such use of the code by competitors -- use which encompasses all of
>> the 
>> copyright permissions of the GPL -- is very open ended and often will
>> not require the competitor to GPL lots of their own code.
>>     
>
> Again this really depends on the code. You cannot make such simple
> generic statements.
>
>   

I can say "often" because I've seen it happen, time and time again.

Yes, it does depend on the code and on the application.   Yes, there
are exceptions.    "Often" is quite fair.


>> It goes further because the company that sold you the software must
>> also permit you to modify it and distribute modifications, subject
>> only
>> to the limits of GPL copyright permissions.
>>     
>
> Correct, but patents works in a different way, and permission to use 1
> claim does not imply permission to use all the patent claims.
>
>
>   



Permission is what permission says.   If you express permission in terms of
copyright rights.... that's what the permission says.

Now, yes, perhaps someone will argue that the GPL covenant is very,
very, weak.   This would be a serious blow to the utility of the GPL.
Worse than simply not solving "the patent problem", it would positively
invite in people to balkanize the corpus of GPLed code by selectively
embodying their patents there.

-t




simo wrote:
One thing is sure anyway. A Patent is often built on many claims, you
are given a covenant only for the claims that apply to the software as
distributed. So if you, by deriving the software start infringing in
other claims, you are not covered by the covenant.
  

The GPL takes all of Mr. Rosen's fun out of leaving the copyright
business and going to the patent business because it requires patent
holders to make a covenant expressed in terms of copyright permissions.

Would GPL allow me to cut an XML lexer out of a web server and
inject it into a database engine?    Clearly that is the intent.   At what
point are related patent claims captured by the GPL covenant and at
what point not?


Sure, they are better off using the GPLed code first and, often, that
will not
mean that they have to GPL all of their own code if they don't want
to.
    

As you know this one really depends on the patent and the code, so you
can;t really generalize.

  

I think you can generalize enough to say this is "very often" the case
and to observe that large-portfolio patent holders who distribute
large quantities of GPLed code will have a very hard time identifying
when they are invoking this consequence.





  
But supposing they haven't been using GPL, and that problem is trivial
to
correct, if the patent holder has been damaged *at all* it isn't by
much.
    

How so? Damages are often calculated based on the amount of money made
by the infringer by infringing the patent. It does not matter if the
same patent was available for free for other uses, the patent holder
decides how to license and make it available, if you loose in defending
your right to use it then damages can well be asked.

  

Google for the "Georgia-Pacific factors" and look for one that talks about things
like "causation" and "but-for".





  
Holders of defensive portfolios, who distribute lots of GPLed
code, are at interesting risk here: their defense stance is greatly
weakened relative to other firms not distributing lots of GPLed
code.
    
        
No, that's wrong again, the patents will not be available to
competitors, unless they use the patent holder GPLed program.
  
      
Which is not much of a restriction.
    

Oh well you can't say that, go ask MS to release the source code of
their OS under the GPL because they infringe on a patent about managing
the Virtual Memory ...

  

No, I mean it the other way.    MS can very often cure the infringement
by releasing much, much less than all of their OS.




  
Right, they must go through the fairly trivial exercise of deriving
the code
which embodies the patent from sources which the patent holder has
conveyed.
    

You assume everything is trivial, I wish it was so, we would have
fantastic software today :-)

  

?

That sort of thing happens all the time.   It's one of the big advantages of
having lots and lots of GPL software.   In my experience (not for patents
per se but just to extract and radically adapt complex components) it's one
of the fastest ways to get code written.





  
And if they have previously failed to do that in some instance, but
the 
problem is trivial to correct, the defensive patent isn't worth much.
    

non sequitur, the infringment was done by non-GPLed and non-derived
software, remedies to past infringement are separate from license for
future distribution.

  

Right.   See "Georgia-Pacific".   What *would* have happened if the infringer
was aware of the issue at the start?   How much extra money did they make
by not using the GPL code in the first place?   How much did the patentee lose?




  
It's interesting that you should bring this up because it applies to
purely 
GPLed programs as well.    Suppose that Acme Co. conveys a Linux
kernel
embodying one of their patents.    I, on the other hand, do a fresh
GPL
implementation of their invention for use in my lisp machine kernel.
If Acme Co. has not themselves conveyed my lisp machine kernel, 
then neither may I.
    

Correct, you will not be covered. What's strange in it?

  

I would say sad, not strange.    This is one reason to like the trend
towards covenants that are broader than the GPL requires.

It's sad because it is against the spirit of software freedoms.   I can
not study one free program, learn from it, and write a new free
program based on what I've learned without risking patent infringement
even though the patents in question are already embodied in the
program I studied.

Are you really going to be happy with a situation where, say,
GNU Awk has a very fast string library but you are prohibited from
using those same algorithms in GNU Classpath?



  
Such use of the code by competitors -- use which encompasses all of
the 
copyright permissions of the GPL -- is very open ended and often will
not require the competitor to GPL lots of their own code.
    

Again this really depends on the code. You cannot make such simple
generic statements.

  

I can say "often" because I've seen it happen, time and time again.

Yes, it does depend on the code and on the application.   Yes, there
are exceptions.    "Often" is quite fair.



  
It goes further because the company that sold you the software must
also permit you to modify it and distribute modifications, subject
only
to the limits of GPL copyright permissions.
    

Correct, but patents works in a different way, and permission to use 1
claim does not imply permission to use all the patent claims.


  



Permission is what permission says.   If you express permission in terms of
copyright rights.... that's what the permission says.

Now, yes, perhaps someone will argue that the GPL covenant is very,
very, weak.   This would be a serious blow to the utility of the GPL.
Worse than simply not solving "the patent problem", it would positively
invite in people to balkanize the corpus of GPLed code by selectively
embodying their patents there.

-t