bug-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 0 of 3] Making PyHurd build again under cython 0.14.1.


From: olafBuddenhagen
Subject: Re: [PATCH 0 of 3] Making PyHurd build again under cython 0.14.1.
Date: Sat, 7 May 2011 18:38:28 +0200
User-agent: Mutt/1.5.21 (2010-09-15)

Hi,

On Mon, Apr 11, 2011 at 11:40:39PM +0200, Arne Babenhauserheide wrote:
> At Fri, 8 Apr 2011 21:05:55 +0200, <olafBuddenhagen@gmx.net> wrote:

> > > It is quite rough (copied a file from cython, needs to be included
> > > cleanly).
> > 
> > What would "included cleanly" mean? And why is it necessary?
> 
> The file is available in Cython and should be used directly from the
> Cython installation instead of being copied in. 

Hm, OK. Then I wonder why you didn't do it properly from the beginning?
:-)

> > I already stated it on IRC, but for the record: I'm not sure this is
> > even possible. The rest of the Hurd can't be upgraded to v3, because
> > of the v2-only Linux code (mostly in pfinet). It all depends on
> > whether the Python bindings can be considered an independant work,
> > or classify as derived...
> 
> That quesion isn't as easy as I'd like it to be. Sending messages to
> Mach should not be viral (shared memory is even used to avoid the
> viral nature of the GPL by corps), but a direct translation from one
> langage to another is a derived work. 

It isn't as simple as that. Whether two pieces of code share a common
address space, is really a technical detail; it doesn't fundamentally
decide about the copyright situation.

Most people assume that linking generally creates a derived work (which
is not always true in fact), and thus distribution in binary form of
components sharing an address space at least always falls under
derivative work. But it's trickier for the source code, which is not yet
linked, but written against the interface of the other components. The
FSF believes that it depends on how "intimate" the interface between the
components in question is. And while it's true that components running
in separate address spaces tend to have less intimate interfaces, this
doesn't always need to be the case.

While a simple exec() call is clearly a trivial interface, complex IPC
as we see it in the Hurd is pretty specific. I'm pretty sure that at
least in some cases, components closely connected through RPC interfaces
would not be considered independent works. But I'm not sure whether this
is true of Hurd interfaces in general, which I'd say fall somewhere in
the middle.

Unfortunately my query to licensing@fsf.org regarding this didn't get
any reply :-( I intended to ping them about it for a long time, but
couldn't get myself to do it so far...

-antrik-



reply via email to

[Prev in Thread] Current Thread [Next in Thread]