'GPL encumbrance problems'

[David G. Miller (aka DaveAtFraud)]

mailinglists at erwinrol.com wrote:

>On Tue, 2006-01-17 at 20:57 -0700, David G. Miller (aka DaveAtFraud)
>wrote:
>  
>
>>> jdow at earthlink.net wrote:
>>> 
>>    
>>
>>>> > From: "Jeff Vian" <jvian10 at charter.net>
>>>> >
>>>      
>>>
>>>>> >> On Tue, 2006-01-17 at 21:40 +0000, Andy Green wrote:
>>>>> >>
>>>>        
>>>>
>
>  
>
>>> Up until recently I worked for a company that developed and marketed a 
>>> closed source, Linux based, network monitoring product 
>>> (http://www.vericept.com).  The company lawyers saw no problem with us 
>>> building and selling a closed source product that included calls to a 
>>> variety of GPLed and LGPLed libraries.  As has been extensively 
>>> discussed over at Groklaw (http://www.groklaw.net) and elsewhere on the 
>>> 'net, headers and interfaces are *not* protectable elements under 
>>> copyright law.  The only way the GPL kicks (via copyright law) in is if 
>>> you actually modify executable GPLed code for your product.  At worst, 
>>> you would then need to make only these changes to the GPLed code 
>>> available as source (there's also nothing that says the maintainer has 
>>> to accept your changes; just publishing them is sufficient).
>>    
>>
>
>The GPL kicks in if you distribute your product. If your program needs a
>GPL library the program is a derived work from that library and so if
>you distribute that program you have to distribute it in a GPL
>compatible way (for example by putting the program under the GPL).
>
>If you have a LGPL library the use of that library is not seen as a
>derived work and hence you are free to choose the license for your
>program. 
>
>  
>
>>> What's good enough for the goose is good enough for the gander so the 
>>> GPL zealots can't have it both ways.  Either implementing libraries that 
>>> are compatible with existing Unix(tm) headers and interfaces violates 
>>> someone's (say SCO's) copyrights or using GPLed headers and interfaces 
>>> to GPLed libraries in proprietary code is legal.  That being said, the 
>>> open source folks have a lot more to lose if headers and interfaces are 
>>> suddenly found to be subject to copyright protection.
>>> 
>>    
>>
>
>That headers aren't protected by copyright doesn't mean the
>implementation of the logic behind those headers isn't protected by
>copyright. If you have a GPL library, copyright doesn't prevent you from
>rewriting (from scratch) that library with compatible headers. Copyright
>does however prevents you from using that GPL library in a way that is
>not compatible with its license.  
>
>So GPL zealots, as you call them, can implement libraries that are
>compatible with other libraries, without violating anybodies copyright.
>And they can put libraries under the GPL and force you to follow that
>GPL when you use their libraries. You, of course, can reimplement those
>libraries too if you want and put them under a different license.
>
>- Erwin
>
>  
>
The point of the goose/gander paragraph is simply linking to someone 
else's code does *not* contaminate your code with their license.  Lots 
of talk and FUD about this but linking doesn't foist a license since the 
headers and interface (which aren't protectable elements) are all that 
is copied.  Just because you program uses a GPL library *does not* 
contaminate it.  As an example, I somehow doubt that an Oracle 
implementation running on Linux makes no use of GPLed library or system 
calls.  Same for VMware and lots of other proprietary software that runs 
on a Linux platform.

Again, IANAL, but I find the argument that the library somehow gets 
"included" in the program to be totally bogus as long as said library is 
a shared object that is still normally distributed.  This argument might 
be made for static libraries or if you literally copy and compile the 
code but that's a different animal.

Cheers,
Dave