CMM support (Was: Re: [Openicc] google SoC starts)

[Graeme Gill]

Craig Ringer wrote:
> I would agree with that, though I'm no lawyer. Specifically, a plugin to
> support the Adobe CMM would not have to be under the GPL because it
> would not be a derivative work of Scribus. We could produce a non-GPL
> (say BSDL) abstract CMM API and implement it against lcms and the Adobe
> CMM. A null/no-op plugin would also not be a bad idea. Scribus's use of
> this API and linkage to the plugins would in no way imply a derivative work.

I'm not sure I agree with this analysis. It's not
necessary that the plugin become a derived work of Scribus,
but rather that what is distributed (which would
include the plugin) is a derived work of Scribus.

If the combined package (work that's distributed) relies
on the plugin for it's total functionality (i.e. it's optional
CMM support), then clearly it's more than "mere aggregation",
and is a derived work of both Scribus and the plugin
(a collective derived work), and the whole is subject to
the GPL as a condition of the permission to copy the GPL
components it's derived from (2a of the GPL).

So if Scribus is GPL, then the plugin "shim" has to
be GPL, and any non-GPL CMM's can't be distributed with
the Scribus package (even if their copyright owners
permitted this) because they add functionality (even if it
is optional), and would be part of the collective derived work.

The reverse also applies. A proprietary application could
implement a compatible CMM plugin architecture, but couldn't
distribute GPL CMM plugins as part of a distribution
package that included the application, because the whole
package would then be a derived work of the GPL components.

[Under most copyright law, even "mere aggregation" forms
  a derived collective work, and only the explicit exception
  in the GPL permits such things.]

Graeme Gill.