Can you alter the MIT license?

[Jules Bean]

On Mon, 15 Nov 1999, Ian Grigg wrote:

> 
> > You can only license something you own the copyright on.
> 
> Precisely.  I'm not sure how you read that you can re-license
> under different terms, this would be tantamount to public domain,
> and the presence of a licence and copyright is evidence that
> the MIT code at least is not in the public domain.

Even public domain isn't your copyright [um.. unless you wrote it, of
course].

Copyright law is based on a 'right'.  The 'right' of authorship pertains
to something which a person contributed original work - and it's sometimes
called copyright.  On a given original work, the copyright may be owned by
several people, if several people have contributed.  Copyrights can be
explicitly assigned to others, bought and sold as commodities.  But they
can't be given up entirely...

'Public domain', literally, applies to something upon which there is no
copyright at all.  This would be something which has no attributable
author, or more likely a long-dead author (copyright thus expired).

However, 'public domain' is often used of software to mean software which
the author has disclaimed all rights on.  This doesn't mean that it isn't
their copyright - it simply means that they have given blanket permission
for anybody to do anything with it.  However, it doesn't mean that someone
else can pick it up and put their conditions on it - since they don't own
the copyright. [Compare the situation of me owning an attractive riverside
meadow.  I put up a big sign saying 'This is a beautiful meadow, and I
make it available to the community.  You may without restriction eat here,
camp here, etc...'.  That, in some sense, puts the meadow in the public
domain, but it doesn't make it any less *my* meadow.  Of course the
analogy is far from perfect, since meadows can't be copied, and software
can be very easily copied...]

> 
> > > - The Free Software Foundation skirted the issue with its repackaging of the
> > > X rendering capability into libxmi.  They licensed the new library under
> > > GPL, but the original source files borrowed from X remain under MIT terms
> > > (even though those files have most likely been modified to embed them in a
> > > GPL library).
> >
> > It's the modification that's the key.  The FSF own the modifications, and
> > the copyright on those is GPL.  MIT own the pre-modified code.  To the
> > extent that the latter can be separated from the former, it can be
> > distributed under MIT terms.  To the extent that they're inseparable, both
> > licenses must be satisfied - possible, since they aren't contradictory.
> 
> The FSF did that?  So that means that you can mix GPL code
> with other code and distribute both under the their respective
> licences?  What effect does this have on that GPL distribution
> clause?

Exactly.  But the effect is that you must satisfy *both* licenses. Now the
MIT license makes no real restrictions on other works its combined into,
but the GPL does.  The problem comes when the GPL demands that you give
certain permissions (the permission to modify and distribute sources) on
any derived work, and this may conflict with one of the other licenses
applicable on a work - but doesn't in the MIT case.

In particular, the GPL says that you must make available the source of the
whole work.  Now this is an additional restriction on top the the MIT one,
but it's not in conflict with the MIT one - it doesn't ask you to do
anything you aren't allowed to do.

> 
> This is a current sticking point for our software *.  We
> have historically been BSD, but are thinking of expanding
> that.  The question has narrowed down somewhat as to whether
> we can include GPL, and the distribution incompatibility
> has complicated acceptance.

Current popular interpretation is that you can't, if your software is
'BSD-with-advertising-clause'.  If you don't have the adv-clause, or
you're happy to give it up, then you can GPL parts of your software, and
include other peoples GPL'ed software.  But you must satisfy the terms of
the GPL (which you are not likely to find a problem).

> 
> Can anyone point to any FSF commentary on how that libxmi
> code was released?

Sorry... I don't know.

>  I'm assuming here that any re-licensing
> of the MIT code wasn't possible as they don't own the code.

Correct.  They can't relicense something they don't own.  

> If I can show clear FSF precedent for mixed distribution
> then acceptance of the GPL would be much easier.

You can easily distribute a combined/derived work of several different
licenses as long as you simultaneously satisfy all the licenses.

You should run all this past your lawyer, of course!

Yours,

Jules Bean
[Not a lawyer; that wasn't legal advice]

/----------------+-------------------------------+---------------------\
|  Jelibean aka  | jules at jellybean.co.uk         |  6 Evelyn Rd	       |
|  Jules aka     | jules at debian.org              |  Richmond, Surrey   |
|  Julian Bean   | jmlb2 at hermes.cam.ac.uk        |  TW9 2TF *UK*       |
+----------------+-------------------------------+---------------------+
|  War doesn't demonstrate who's right... just who's left.             |
|  When privacy is outlawed... only the outlaws have privacy.          |
\----------------------------------------------------------------------/