Ben Finney writes ("Re: ad hoc license: is it DFSG-conformant ?"):
> Ian Jackson <ijack...@chiark.greenend.org.uk> writes:
> > Knoppix needs:
> >
> > - permission from upstream for Knoppix modify and redistribute the
> > modified versions. This is granted directly to Knoppix by
> > upstream in thier licence.
>
> I think upstream don't have the exclusive power to grant license to
> Knoppix in the modified versions; that's for the copyright holder in
> those modified versions.
For a derived work, copyright law requires, simply, permission from
all the copyrightolders.
> Let's say that a Debian recipient, Carol, modifies the work such that
> they now also hold copyright in the modified work.
>
> Now both upstream and Carol hold copyright in the modified work.
Yes.
> As you say, upstream have granted Knoppix license in the work, but
> that's in the work prior to Carol's modifications. Knoppix needs license
> in the work as modified by Carol, and so needs license granted by Carol
> also.
Carol must grant permission. Typically this is done by Carol writing
Signed-off-by in a git commit, or adding her name to the list of
copyrightholders about the licence notice, or some such. If Carol is
the Debian maintainer she'll probably grant permission by writing
something in debian/copyright, if nowhere else.
> So how does Carol have permission to extend that license on the modified
> work? What in the grant Carol received lets her do that?
I don't understand what you mean by "extend" the licence. Carol
doesn't need to "extend" the upstream copyrightholder's licence. A
licence is simply permission. Permission can be granted to do any
things or subset of things that can be sensibly described in writing.
Because a licence is simply permission.
The upstream copyrightholder has already granted permission to anyone
to distribute modified versions.
Concretely: Knoppix want to distribute Knoppix's version. Knoppix,
Carol, and upstream, all have copyright in that.
Knoppix don't need their own permission, obviously. (But they will
presumably grant the same licence to everyone, as is conventional.)
They need Carol's permission. Carol gives that permission by adding
her name to the appropriate file(s) where the copyrightholders grant
permission.
They also need upstream's permission. Upstream have already given
that permission (specifically, permission to distributed modified
versions, such as the version produced by Carol, and that subsequently
produced by Knoppix). That permission is once again a public licence
- ie, permission is granted to Carol, Knoppix, and anyone else.
> Mind you, I note the same issue in the BSD 3-clause and 2-clause texts
> (which combine grant of license, with license conditions). Unlike
> typical grants for GPLv3, Apache License 2.0, and Expat license grants,
> the BSD licenses don't explicitly grant permission to redistribute under
> the same license.
The BSD licences grant permission to redistribute and simply do not
restrict what licence a downstream grants.
> So I'm grateful for the concrete example, and am re-assessing my
> position; it's more likely my conclusion is incorrect. I would like to
> know exactly where — i.e. what in my above reasoning is incorrect.
I hope this has helped.
Ian.
(not a lawyer, but a fairly frequent litigant)
