Re: Summary : ocaml, QPL and the DFSG.

[David Nusinow <david_nusinow@verizon.net>]

On Thu, Jul 22, 2004 at 02:07:54PM -0400, Brian Thomas Sniffen wrote:
> David Nusinow <david_nusinow@verizon.net> writes:
> No, you don't have to find one.  Just write a very, very simple one.
> I don't think it can be done in a free way, but if you show me one,
> then I'll believe you.

I've thought about this for a while, and I think that perhaps the simplest way
that it would work would be to distribute changes to your immediate upstream,
rather than the original upstream. You would have to have some relationship
with your immediate upstream in order to get the software to modify in the
first place, so there should be no additional fee associated with distribution
upstream in this case. 

> >> I do think Sven might disagree, and have reason to be just a little
> >> testy that I've made spamming him a condition of distributing
> >> modifications to my software.  If Linux were licensed that way, Debian
> >> would have to send one kernel source tree per download per kernel
> >> copyright holder to poor Sven.  That would be thousands of kernel
> >> sources.  Surely, enough to put debian.org and its mirrors into some
> >> unhappy territory.
> > Ok, I misunderstood your question. I assumed Sven would want the changes. This
> > would classify as discrimination against Sven, and would fail the DFSG.
> > Fortunately, this is not necessarily the case with forced upstream distribution
> > clauses.
> Why is this discrimination against him?  I think that's fairly
> contorted, in comparison to the simplicity of saying that a Free
> license cannot compel me to initiate action, only to do some things in
> particular ways, and this compels me to initiate communication with somebody.

I didn't realize you were headed in the "must pet a cat" direction with this.
Harassing Sven when he doesn't want it would be discrimination. I'd rather not
justify this, since I think it's pretty self-evident. As for the "must pet a
cat" side of the question, I agree.

> I asked if a free license could, in your view, require that any time
> you distribute modifications, you also send a copy to the original
> author.  That would require sending them on *every download*.  If
> you're a mass distribution site, that's a problem.

Ok, now I see. I'd see this as failing DFSG 1 because it would effectively
prevent Debian from distributing the software. Not even getting in to fees and
whatnot, this would definitely be a real restriction. Even looking at this from
a pragmatic, rather than ideological point of view, this holds.

> >> >> Can I say you must do it by a non-digital mechanism?
> >> > This question could be asked for forced downstream source distribution as well.
> >> > Why not?
> >> Because those are expensive.  Real mail costs a lot more than e-mail.
> >
> > Ok then, since this would fail the fee test by my definition of the word fee.
> > But fortunately I've never seen a forced upstream distribution clause with this
> > requirement, which would make it non-free.
> 
> OK; but requiring me to use my network connection is not a fee, not
> even if I pay by the bit?  Is there a bright line, here?  Or just a
> vague idea that some costs are large enough to be non-free, but very
> small costs are not worth worrying about?

There's definitely no bright line, but again, what if the license forced you to
send your changes downstream in hard copy, rather than upstream? I'd say that
this would fail DFSG 1 too.

> >> >> Can I say you must sign your changes?
> >> > As above, this could be applied to downstream distribution. Why not, given the
> >> > DFSG? (The dictator test obviously would apply, but I don't know if I agree
> >> > with it as a functional tool)
> >> Because it compels me to reveal my identity to distribute changes,
> >> which is a cost.
> >
> > I don't consider this a valid argument. You reveal your identity distributing
> > changes downstream as well.
> 
> No I don't.  I can drop CDs in the street, or paint code on walls.

Ok, let's invoke the pet the cat argument from earlier. You said above "in
comparison to the simplicity of saying that a Free license cannot compel me to 
initiate action, only to do some things in particular ways". This forces you to
distribute your changes in a particular way. The revelation of your identity is
the way in which you must distribute your changes. It's a restriction on
behavior, much like the GPL places other restrictions on distribution of
changes.

> > Furthermore, nowhere in the DFSG is privacy guaranteed (and I won't
> > accept discrimination as a valid reason for this because the license
> > is not written with the intent to discriminate against people who
> > need to keep their identity secret).
> >
> >> >> Can I require a license from you?  More free than otherwise compelled
> >> >> by the copyleft?  What about a non-free license, can I require that?
> >> > No, because this obviously fails DFSG 7.
> >> No it doesn't.  My license passes on to them.  It's just that your
> >> changes have to be under a more or less restrictive license.  Ah, you
> >> mean DFSG 3.
> >
> > No, I meant 7, but 3 applies as well, thanks for clarifying for me.
> 
> I don't see how requiring that you license your material under a more
> or less free license than I used when giving the material to you fails
> DFSG 7.  Can you explain that to me in more detail?

"The rights attached to the program must apply to all to whom the program is
redistributed without the need for execution of an additional license by those
parties."

I read this as being very similar to DFSG 3. If you place a more restrictive
license for those you distribute your modified program to, they lose the rights
that they have. Further, if upstream forces you to give them a license that's
non-free then the rights attached to the program don't apply to you and them
equally. Are there holes in this interpretation?

> For example, the QPL is non-free because I can't distribute to the
> initial author under the same copyleft he used to distribute to me.
> It fails DFSG 3.  It is also non-free because it compels me to retain
> access to anything I link to the licensed work forever, if I ever
> distribute it, thus imposing a fee on distribution and violating DFSG
> 1.

The first point I definitely agree with. Not entirely sure about the second
yet, mainly due to lack of clarity with respect to timeframe, but it's moot
since the first makes the QPL non-free in my book.

> > I don't believe that forced upstream distribution is necessarily free mind you,
> > just that the extringent requirements in the actual license need to be taken in
> > to account, which is what I meant by "level of detail" in an earlier mail.
> > Ultimately, I think the Desert Island Test needs refinement, because as it is,
> > it strikes me as rather crude.
> 
> I think that's a different but worthy discussion.

I'm glad you think so. I'll try and come up with some ways to refine it,
although it's been difficult going so far.

 - David Nusinow