Re: licence issues

[Xavier Leroy <Xavier.Leroy@inria.fr>]

> at the moment there is a thread on "comp.lang.functional" discussing
> legal aspects of the OCAML-distribution policy.
> Some people believe it is too restrictive and they thus rule it out for
> their purposes. But I think this is mainly due to misunderstandings of
> the licence and/or that the licence is not always explicit enough.

The point of the license that is discussed on comp.lang.functional is
how modified versions of OCaml can be distributed.

What the INRIA license states (not very clearly, I agree), is that
public distribution of modified versions is allowed in the form of
source diffs only.  Other forms of distribution (in particular as
precompiled binaries) require prior authorization from INRIA.

The intent of this clause is to prevent "unfair" reuse of OCaml code
in projects that could be harmful to the whole OCaml effort. e.g. in
competitors' projects.  For instance, I believe one could make a
killer Java compiler by taking the OCaml native-code generators and
garbage collector and bolt them onto a Java front-end.  We feel this
would be an unfair use of the OCaml sources, and would compromise our
research effort in the field of functional programming.

A license such as the GPL doesn't address this concern: it would allow
this unfair use as long as the modified version is distributed under
the GPL and with source code.  This is better than distribution
without source code at all, but is still damageful to us.

By requiring distribution as source diffs, we force users of modified
versions to download the OCaml source code and apply the patches
themselves.  This way, they are made sufficiently aware that what they
are using is nothing but a modified version of OCaml.  If we allowed
the distribution of precompiled binaries, this would be no longer the
case.

My personal feelings is that this point of the license is a bit too
restrictive.  However, the alternatives we know of are not restrictive
enough: the classic free licenses (GPL, BSD, etc) don't protect the
authors at all against unfair reuse.  The INRIA license errs on the
conservative side, that's all.  If we could find something less
restrictive but still protecting us against unfair reuse, we would
gladly change the license.

It should be pointed out that the current license has (as far as I
know) never prevented any reasonable use of OCaml.  As a case in point,
Debian has an OCaml binary package because they didn't have to modify
anything in the source to make their package.  Similarly, the license
didn't hamper the development of OLabl, which is clearly a derivative
work.  Finally, INRIA has never refused any requests for license
exemptions that have been submitted in the past, and there are no
indications that this will change in the future.

> It would be a real pity if people ran away /
> didn't look at OCAML due to some unclear licencing issues.

It's hard to please everyone.  The GPL makes some other people run
away (mostly industrial users who don't want to release the source for
their modifications).  My hope is that while the current license might
drive away a few license ayatollahs, it should not be an obstacle to 
all other OCaml users.

- Xavier Leroy