Public Domain Python

[Tim Peters]

[Huaiyu Zhu]
> I have not much problem with the licence per se than the fact that it
> appears to be imposed forcefully and retrospectively.  You may
> say that they always had the right to do so,

Of course the copyright holder has the right to change licenses.  GPL, CWI,
X/MIT, BSD, ... doesn't matter, that's always the copyright holder's
perogative.  Unsure what "retrospectively" means.

> and that the change is for good,

That I don't know.  I do know that some of the Python Consortium members,
who build businesses in whole or in part on Python, and who think a lot more
about lawsuits than Guido & I combined ever would over 1000 lifetimes, do
seem to think it's an improvement.  Certainly the *way* in which CNRI did
this was a public relations disaster (although I doubt they're aware of
that!).

> etc, but any possibility of such things happening is disturbing.

Indeed it is.  That's why I favor public domain, or assigning copyright to
an independent entity who promises *not* to change the license at whim (the
FSF is one such entity), or at least not without serious community
involvement (such as the ASF wrt the Apache license).

> ...
> I still can't figure out the answer to this simple question: Is
> it legal for anyone not employed by CNRI to pull the Python CVS at the
> point right before the new CNRI licence was released and distribute it
> with CWI licence?

"Simple" and "legal" just don't go together.  Get over it <wink>.  The laws
here were written long ago without any concept of software in mind, and
their application *to* software is still highly debatable.

As I said, we have a legal opinion saying "fine! -- probably".  There's an
irony here for *you* to ponder, though:  since the GPL is grounded solely in
copyright law, then had Python been under the GPL, the question of whether
something merely checked into a CVS tree is also under the GPL would quite
possibly rest on whether "checked into a CVS tree" meets the technical
meaning of "publication" under copyright law.  I'm almost certain CNRI would
say it does not, and I expect they'd have a very strong case.  The GPL lets
you make all the private modifications you want without falling under its
terms (it has to -- it would have to go outside copyright law to claim that
you couldn't):  it's not until you *distribute* that copyleft kicks in.  So
does checking into a CVS tree constitute distribution?  Try to think like a
lawyer-- or even like Usenet poster not in love with their own position
<wink> --and you'll find plenty debatable there.  So it's certainly not
stronger than the CWI license in *that* respect.

> ...
> I thought the CWI licence was liberal enough to allow puting into public
> domain - but maybe that's too much.

Absolutely not.  Only a copyright holder can reliniquish their copyright
claim.  You could certainly put your *own* enhancements to Python in the
public domain, but that can't affect what you built on.  And, to the
contrary, the CWI license specifically requires that you retain their
copyright notice on their work (else you've violated the terms of the
license, and have lost *all* rights to use their code).

> ...
> With GPL, as long as it is authorized, no one can take it back.  That's
> sufficiently clear from my pov.

Sure.  What isn't clear is why you think the GPL is special in that respect.

> ...
> Then what was the reason they give for having the right to impose the new
> licence?

They're the primary copyright holder.  It *is* their right.  Just as it's
your right to change the license on anything you hold copyright (GPL'ed or
not).

> If the only problem was Guido's employment obligations, what prevent
> you from releasing it under CWI, for example?

Well, I work for Guido, so, in a sense, his obligations became mine.

> Ok, maybe you don't want to - but you could, right?

It's my belief that we could, yes.  It may or may not have ended up in court
if we tried.  The mere *possibility* of ending up in court was in fact a
very strong deterrent against trying (have any idea how long court battles
can drag on in this country?  how expensive they are?  I do -- but I
personally still thought it was a good bet).

>> Again, if you had read the licenses, you would know that CNRI's
>> name is in every paragraph of "the CWI" license.  The CNRI license
>> *FAQ* implies that CNRI's name was only in the liability disclaimer
>> portion of the CWI license, but it only takes 10 seconds to verify
>> that it's everywhere else too.  Like I said, this is Usenet <wink>.

> However, they are not displayed in the command line.  Here're my
> old copies:
>
> $ /usr/bin/python
> Python 1.5.2 (#1, Sep 17 1999, 20:15:36)  [GCC egcs-2.91.66 19990314/Linux
> (egcs- on linux-i386
> Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam
>
> $ python
> Python 1.6a2 (#2, May 10 2000, 16:58:36)  [GCC 2.95.2 19991024
> (release)] on
> sunos5
> Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam

I see that as no more than a curiosity.  The Python web site even pointed
this out, and reminded people that copyright was held regardless of whether
it was explicitly stated (which is true).  In this case, it may be no more
than that CNRI's attorney never ran Python so never noticed this.  They
*did* notice the CWI license, though, and Guido put CNRI's name all over it
at CNRI's attorney's direction.

> Anyway, all this just add to the argument: if the earlier releases were
> authorized and under GPL, it would not matter what a later new licence is.

Surprise:  I still see nothing special about the GPL in this respect --
unless copyright had also been assigned to the FSF.  To the contrary, due to
some of the points above, I think it would have been weaker than the CWI
license wrt the status of stuff in the CVS tree.

a-license-written-by-a-programmer-is-likely-to-be-too-clever-for-its-
    own-good(indeed-as-rms-says-about-larry-wall's-artistic-license)-ly
    y'rs  - tim