Subject: Re: BSD-like licenses and the OSI approval process
From: "Ben Tilly" <btilly@gmail.com>
Date: Thu, 18 Oct 2007 11:25:33 -0700

 Thu, 18 Oct 2007 11:25:33 -0700
On 10/17/07, Philippe Verdy <verdy p@wanadoo.fr> wrote:
> Ben Tilly [mailto:btilly@gmail.com] wrote:
> > Envoyé: mercredi 17 octobre 2007 21:56
> > À: Chris Travers
> > Cc: License Discuss
> > Objet: Re: BSD-like licenses and the OSI approval process
> >
> > On 10/17/07, Chris Travers <chris.travers@gmail.com> wrote:
> > > On 10/16/07, Philippe Verdy <verdy p@wanadoo.fr> wrote:
> > > >
> > > > So if the fact that they use "open source" terms for describing the
> > licence
> > > > does not matter. We are ONLY interested in the compatibility of the
> > licence
> > > > with our set of approved licences. And for me this is just enough,
> > however
> > > > IANAL...
> > >
> > > However, taking this viewpoint will effectively prevent the OSI from
> > > asserting that non-OSI approved licenses should not be called open
> > > source (as Mr Tiemann has argued in his blog).
> >
> > Not so fast.
> >
> > IANAL either, but it seems to me that if the copyright holder says,
> > "My license is open source because it grants sufficient permissions to
> > allow it to be relicensed as X", that statement itself is sufficient
> > permission to allow someone to distribute under the terms of X.
> > Therefore the software is now implicitly licensed as X and is
> > effectively an open source license.
>
> You don't understand: I mean that if a software writer wants to designate
> its software as open source, he can do so without any legal consequence.
> The only thing that is legally important is not the fact that its licence is
> said to be "open source" (without any reference to OSI), but the fact that
> the licence *asserts* as being compatible with another explicitly named
> licence that you consider being "open source" (BSDL, MPL, GPL...).

If that was what you meant, you could have expressed it far more clearly.

I agree that any license that explicitly gives permissions to apply an
OSI approved license is trivially open source.  To the best of my
knowledge, nobody has a problem with this.  The problem are all of the
permissive licenses that *don't*explicitly say that.  People using
that software claim to be open source.  Based on the facts, they are
obviously right.  For a variety of reasons the OSI doesn't want to
approve all of them individually.  And the OSI doesn't like having
people using non-OSI approved licenses and calling them open source.

> One way a licence can do that is by explicitly granting the user to choose
> between either the terms of the licence itself, or the terms of another
> licence, including if the licence requires (in that case) to terminate the
> licence at the same time as he chooses the alternate licence.

The permissive licenses that are at issue, like the Postgres license,
do not include such terms so that doesn't help.

> This is not like multiple licencing, where all licences apply
> simultaneously, because the text clearly requires the user to choose
> irrevocably between the two texts, and only one licence applies at any time.
>
> I won't say the same thing about the term "copyleft" which was uniquely and
> first defined by the FSF as necessarily meaning "GPL compatible", and
> there's apparently no precedent (so the "copyleft" term is probably
> defendable in a court by free software supporters, but IANAL to assert this
> would be the case if "copyleft" has not be registered as a trademark by the
> FSF, allowing it to forbid its usage by some licensor whose licence is not
> compatible with the GPL).

Um, copyleft does not mean "GPL compatible".  It means a license whose
copyright terms work to make sure that all recipients will have the
ability to do things that would normally be forbidden by copyright.
(ie make, modify and distribute copies.)  The GPL merely happens to be
the best known copyleft license.  There are others, but they are less
important because people don't use them much.  For example the Design
Science License is a copyleft license that is not GPL compatible.  See
http://www.gnu.org/licenses/dsl.html for the terms.

> I absolutely don't matter in fact about the terms. What I need and want is
> ONLY making sure that the licence is compatible with some other well known
> "open source" or "free" licence. And the job of the FSF and of the OSI is to
> study if such licences are compatible with their rules.
>
> Unfortunately, there exists licences that are considered compatible with OSI
> rules (and approved by OSI) which are considered NOT compatible with the GPL
> by the FSF (despite the GPL itself is compatible with OSI rules). This means
> that the licence compatibility is NOT a transitive property, but only a
> PARTIAL ORDER (like in a oriented graph, where NO shortcut is implicitly
> allowed):

Poetic justice.  You tried to say a very simple thing in a very
complicated way, and got it hopelessly wrong.

First of all transitive means that x < y and y < z implies that x < z
for all x, y and z.  Partial orders are always transitive.  Now what
you said about a random license x, the OSI y and the GPL z is
equivalent to x < y and z < y does not always mean x < z therefore <
is not transitive.  Um, not so.

Incidentally license compatibility *is* a transitive relationship.
Meaning that if license X grants sufficient permissions to allow
incorporation in a codebase under Y, and Y grants sufficient
permissions to allow incorporation in a codebase under Z, then X
grants sufficient permissions to allow incorporation in a codebase
under Z.  (Trivially so by incorporating X in a codebase under Y, then
incorporating that codebase into one under Z.  Now stuff under X was
incorporated under Z.)  I am assuming here, of course, the proposition
that "sufficient permissions to allow incorporation" for a given pair
of licenses has a definitive yes/no answer.  And yes, I'm aware that
this assumption can be debated.

Also I should note that license compatibility is not a partial order.
There are examples of permissive licenses X and Y such that X and Y
are different (they have different text), but X can be incorporated
into Y, and Y can be incorporated into X.

Of course all of this theory is irrelevant to the OSI's problem,
because the real issue is that it is not always obvious looking at a
license whether it is compatible with another, and it is particularly
not always obvious whether it fits OSI's definition of open source.
(If it was obvious, then there would be no need for the OSI to certify
specific licenses as being open source.)

> If licence A is said compatible with licence B which itself is said
> compatible with licence C, this does not imply that licence A is compatible
> with licence C. It does not imply also any reverse order.

I agree that compatibility is often one way.  But the rest of your
assertion does not seem correct.  Certainly it isn't with the
definition of compatibility that I used.

Could you explain what you mean by compatibility?

> This really complicated the things, and that's why avoiding the
> proliferation of licences is highly wanted as it really fragments our
> desired commons. And one way to reconcile things and reconsolidate this
> licencing graph is by making licences that are EXPLICITLY said (in their OWN
> text) to be compatible with several other wellknown licenses, or for
> free/open source software makers to avoid writing their own licence by
> providing their software with multiple licences (and providing a way to
> contact them for getting supplementary licences if one of those proposed
> licences are not satisfying for users, due to the difficulties of
> integrating it with other software covered by one of the proliferating other
> licences).

Great, our solution to people sticking with a plethora of permissive
licenses that are not OSI approved is to convince them all to change
their licenses to ones which are still not OSI approved but which say
you can relicense it under an OSI approved license?

If we knew how to convince them to do that, why couldn't we just
convince them to use an OSI approved license instead?

Ben