RE: [oc] New License update

[bob.cousins@philips.com]

>But I don't know what the GPL means in the context of chips.

Ok, fair point ;-)

I must make the caveat that I don't have detailed knowledge of logic synthesis
so I may be talking nonsense.

I accept that there needs to be a specific licence made for hardware because
the details are different.
 
I am sure companies could find legal loopholes in the GPL, is there a legal
definition of link? Linking effectively means putting into the same executable 
file, the equivalent would be putting stuff onto the same package (i.e. the final 
thing that is distributed to the customer).

Isn't logic synthesis analogous with compiling? Assuming your GPL
source is in C, how you convert this to machine code is between you and 
your compiler. You can then sell (or whatever) the implementation for CPU X without
needing to reveal details of your compiler. 

The difference I see is that typically you would not patent your compiler, so
anyone else can produce an implementation for CPU X using their own compiler.
There is no issue with competing implementations.
However a fab might be tempted to patent its /implementation method/ which 
might restrict others from producing their own implementaton.

Similarly if you have a proprietary RTL there is no problem bundling that up
with LGPL code and selling the implementation. The crux is that any changes
made at the /source/ level to the LGPL or GPL code need to be distributed.

My concern is really that a fabber could change an include file in the VHDL to 
include his library, and at the same time make other, possibly generic. useful 
optimisations to the design, and then call the whole thing his, and at the same
time prevent anyone else producing competing chips.

-- 
Bob Cousins
Software Engineer.
"Beginning and end are part of a single ring and no one can comprehend
its principle. " - Chuang Tzu