[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

GPLv3 / Affero / RPSL

On Mon, Mar 10, 2003 at 05:17:39PM +1000, Anthony Towns wrote:
> > But free software was never about forcing people to contribute back.  
> No, but the GPL is about forcing people to pass the freedoms they have
> onto their users.
> Note that you do _not_ get to assume "privacy is good and moral and a
> right of both individuals and corporations". Justify it in other terms,

Since nobody else seems interested in bothering to try this, I guess I'll
have to flip-flop again.

So, the key point in my suggestion that a license could require you to
send changes to the original author on request, was that you really need
to limit the number of copies you can potentially end up distributing,
or else you'll end up having to setup a full distribution mechanism,
which would be overly burdensome and hence non-free.

However, what that means is you need to decide on an "author" and list
their name in the license. But if you then have two different programs
that both do this, and choose different names, you're left with a
quandry. Do you just choose one -- in which case, there's nothing stopping
you from putting your own name in their, or the name of a friend and doing
whatever you want -- or do you put both in, in which case you end up with
having to reply to anyone who's ever hacked on the code, which in the case
of the linux kernel, eg, is probably enough people to be burdensome. The
other option, of course, is to just say you can't combine code from the
two projects.

I think that pretty much makes it unacceptable for GPLv3 (it ends up
being useless, non-free or incompatible with itself respectfully --
all of which are unacceptable in the premier copyleft license).

I think we have two sorts of free licenses. One set, which includes BSD
and GPL licenses, which basically give users and authors the same rights;
and the other set, which includes the QPL and licenses with patch clauses,
which give the original author special consideration. I can't see any
reason not to put licenses that require sending copies to the author
into the second category -- free, but deprecated.

The general case seems to be this. If you've set up a web app, people with
an interest in the source code are in the following classes:

	(a) yourself
	(b) people you specifically choose to let run your web app themselves
	(c) your users
	(d) the original authors
	(e) the general public

Requirements about distributing source code to (a) and (b) are largely
trivial, and not worth discussing. Requiring you to distribute to (c) and
(e) creates a large burden on you, that's roughly equivalent to something
like "If you use this program, you're required to donate $5 to the FSF".
Meanwhile, distributing to (d) can be reasonably managed in _very_
restricted senses (your costs are covered, only by request, and only to
an individual representative of the authors, not any of them who ask).

Which is to say that I think it's fundamentally impossible for the GPL
to close this loophole.


Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

  ``Dear Anthony Towns: [...] Congratulations -- 
        you are now certified as a Red Hat Certified Engineer!''

Attachment: pgp0LDrr4InAb.pgp
Description: PGP signature

Reply to: