[pypy-dev] speed.pypy.org and mercurial

Ben hyarion at iinet.net.au
Sat Dec 18 12:10:11 CET 2010

On 18/12/10 8:11 PM, Paolo Giarrusso wrote:
> On Fri, Dec 17, 2010 at 16:06,<exarkun at twistedmatrix.com>  wrote:
>> On 02:48 pm,tobami at googlemail.com  wrote:
>>> @Jean-Paul:  I thought it would only be a problem if the license was
>>> GPLv3?
>> Yes, that's right.  But I got sad anyway.
>>> But in any case, what is the conclusion, if any? Can the Mercurial API
>>> be used from LGPL code or not?
>> This seems unanswerable.  If you believe some people, yes.  If you
>> believe other people, no.  The only way to resolve the conflict is to do
>> it, and either get sued by the copyright holder or not get sued by the
>> copyright holder (you could argue that the outcome of the lawsuit must
>> also be considered, but by the time you've gotten sued you probably
>> don't care anymore).
> Well, one point is that the matter is complex enough to make you
> suable. Moreover, the point that you're technically not linking seems
> extremely technical and specific to Python, and even to CPython - a
> JIT compiler links together the code, even more than it happens for C
> libraries. Therefore, explaining it to a judge would take an extremely
> expensive lawyer.

The GPL doesn't mention linking at all. It applies to "the Program or 
any derivative work under copyright law: that is to say, a work 
containing the Program or a portion of it, either verbatim or with 
modifications and/or translated into another language". Linking is only 
relevant because traditional static linking causes parts of the library 
to be included in compiled binaries, which automatically makes them 
derivative works.

Downstream recipients of your code are also perfectly within their 
rights to combine GPL code with code released under an incompatible 
license; provided they never distribute such code, the GPL imposes no 
restrictions. So the fact that the JIT compiler links code together on 
the end user's system is not relevant. The GPL terms only come into play 
if the CodeSpeed source can be considered a "derivative work under 
copyright law" of Mercurial.

If CodeSpeed is designed to be support several version control systems 
via independently installable modules, then I don't see how it could be 
considered a derivative work of Mercurial. The integration module might 
be, but to accept that the existance of such a module makes CodeSpeed a 
derivative work means that CodeSpeed is also a derivative work of every 
version control system, or at least will be as soon as a module is 
released for each system (which could be done by other people). I find 
that rather ludicrous.

That said, I'm not a lawyer, and I'm applying common sense reasoning 
here, not knowledge of how the legal system actually treats things.

-- Ben
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/pypy-dev/attachments/20101218/d5e1e1a7/attachment.html>

More information about the Pypy-dev mailing list