License Committee Report for July 2007

[Wilson, Andrew]

John Cowan wrote:

> It's not as bad as you think.  There are two kinds of derivative works
> made from LGPLed code: in LGPLv2-speak, they are "works based on the
> library" and "works that use the library".  The corresponding LGPLv3
terms
> are "works based on the Program" and "Combined Works", respectively.
> 
> The compatibility matrix you speak of refers only to the first kind:
you
> cannot add a LGPLv3-only module to an LGPLv2-only library.  However,
you
> are entirely free to make use of an LGPLv2 or LGPLv3 or
LGPLv2-or-later
> library from any kind of program, whether under an FSF license,
another
> FLOSS license, or a proprietary license, provided you obey the basic
> constraints (allowing the user to replace the library with new
versions,
> basically).

Beg to differ, it is as bad as I think (or, that is to say, worse than
you think).  Look specifically at the case of a GPLv2-only program
linking to an LGPLv3 library where the "system library" exception does
not apply.  GPLv2 says the entire Combined Work
falls under its copyleft.  This is trivially easy with an LGPLv2
library, since the library can be automatically relicensed as GPLv2.
On the other hand, the LGPLv3 library is in fact GPLv3 with extra
permissions which may be stripped away to leave bare GPLv3 ... which
wants the entire Combined Work to be licensed under *its* copyleft,
which
FSF says is incompatible with GPLv2's.  Conflict of licenses ensues.

Now look at the case of linking a proprietary program with the same
LGPLv3 library.  Since one of those additional permissions is to bound
the scope of v3 copyleft from spreading to the entire Combined Work,
no conflict of licenses.

I'll take those ibuprofen tablets now....

Andy Wilson
Intel open source technology center