Re: licence issues

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

> [Hypothetical Java compiler reusing the OCaml back-end]
>    You are quite right that you have made something special and
> valuable.  And if the licence is very free, then there is a good
> chance that other people will find your code base useful in the sort
> of way you suggest.  If that happens, you will have made an even
> bigger contribution to the public good than you have already

I agree that in the long term reuse of free source code is generally
beneficial to the public as a whole, and it's definitely the goal of
government-funded research to offer its results to everyone.  (Also, if
we were that afraid of baddies stealing our ideas, we wouldn't
distribute the source code at all.)

However, there are two problems with the argument above.

First, it assumes that the reuse of OCaml code cannot harm the development
of OCaml itself.  But this is not necessarily so.  In the silly
scenario I mentioned, I can already hear INRIA's management telling me
"Ok, you told us that Caml and functional programming and type
inference and so on are good. We believed you. We supported you during
many years. Now the main outcome of your effort is a Java compiler
developed by others who get all the credit for it. And you still want
us to support you ?"  All these long-term benefits of open source are
not going to make much good if we get killed in the short term.

> , and you
> will, whatever happens, get a significant portion of the credit.

This is the other problem.  My experience is that this is not always so.
For instance, how many users of Moscow ML know that the compiler and
runtime system come straight from Caml Light?  The MoML authors did
everything right (putting copyright notices on all files,
acknowledging INRIA in their Readme files).  Still, I don't think we
got any credit (in the general sense) from this code reuse.

This is a very touchy topic for research.  Research lives off peer and
public recognition, just like commercial software lives off sales.

>    This is the standard theory of open source economics and I think
> it's mostly true.

I'm familiar with that theory.  By the way, my yougest baby,
LinuxThreads, is under the LGPL.  But it's not a research project,
just a hobby.  The open source economics don't take into account a
number of factors that are important for research, such as:

- The need to maintain some competitive edge in order to justify one's
existence (research that doesn't have any edge over what others are
doing is a contradiction in terms);

- The need to get exact credit for one's work (the system of journals
and conferences does it quite well for research papers; no equivalent
system exists for source code yet).

I wish open-source "ayatollahs" (as I called them before) could
think about these issues rather than just bullying everything that
is not GPL.  (The latter takes a lot less thinking, of course.)

All the best,

- Xavier Leroy