Re: Bug#227159: ocaml: Worse, the QPL is not DFSG-free
Matthew Garrett <mgarrett@chiark.greenend.org.uk> writes:
> Brian T. Sniffen wrote:
>>Matthew Garrett <mgarrett@chiark.greenend.org.uk> writes:
>>> I think you would find it hard to gain consensus that "fee" should be
>>> interpreted in this manner. The GPL requires that I provide either all
>>> the source code, a written offer or alternatively information regarding
>>> the written offer I obtained. A reading of "fee" that broad would cause
>>> the GPL to fail on the same grounds.
>>
>>Not at all. The GPL lets me put together (binary+source) bundles and
>>distribute those -- this is a Free license. The other paths,
>>involving written offers, don't have to get involved. QPL says that
>>if I distribute (modified binary+source) to you, then the initial
>>author can come along *any time he likes* and demand a copy from me,
>>together with a license to modify and distribute it.
>
> What if I only wish to distribute binaries? The requirement that I
> distribute source alongside them is a fee. It's not necessarily one that
> applies to Debian, but it's one that applies to our users.
No, because you can't distribute binaries or source at all without the license.
So saying "you can distribute the class of derived works
binary+source" is fine; saying that you can distribute binaries+source
and must also send binaries and a license to the initial author is not.
>>The requirement that I give a copy to the initial author, just because
>>I'm sharing it with a friend, is non-free.
>
> Why? Asserting this doesn't make it true.
Well, for a start because it compels me to keep a copy forever. It
also compels me to send a potentially huge number of messages. For
example, it doesn't say I only have to send him one today. Tomorrow,
he can demand another. Maybe he used up the first. Maybe he lost
it. Doesn't matter, I have to send another.
In any case, even *one* distribution to a person I don't want to
distribute to is non-free.
>>>>Additionally, 6b requires that I license my modifications to others
>>>>under a *more* permissive license than the QPL. Those to whom I give
>>>>my items (presumably meaning my modifications) must be licensed to
>>>>distribute modified copies without charge, and the QPL imposes a
>>>>charge. Since I can't distribute my modifications under the same terms
>>>>as the license of the original software, this also fails DFSG #3.
>>>
>>> I think that it would be even harder to claim that "charge" covers this.
>>
>>But this isn't about a charge -- this is about whether I can
>>distribute my changes under the same license, the QPL.
>
> Christ. Yes, it is entirely about a charge because the QPL explicitly
> talks about charges. In order to claim that you can't distribute under
> the same terms you need to demonstrate that the requirement to provide a
> copy of the modifications upstream equates to a charge. I've seen no
> convincing arguments of the sort.
Oh, I'm sorry -- I thought that was clear from the first part. You
don't think that an agreement to keep every modification I ever
distribute to anybody around forever in case the original author wants
a copy is a charge?
>>This isn't about the success of lawsuits; it's about complying with
>>the license as written.
>
> Why is it more important to comply with the license as written than it
> is to believe that we can provide freedoms to our users without them
> being sued?
>
>>As to the dissident example, China and the US are working together on
>>copyright issues while opposing each other on political issues. If
>>China, say, buys a QPL author and arranges for him to publish a
>>demand, they can then enlist the US' help... of course this scenario
>>is convoluted.
>
> I'm sorry, I don't understand this at all.
>
>>The real reason for the dissident example is this: any requirement for
>>forced distribution, for action where there would otherwise be total
>>inaction, is non-free.
>
> Why?
>
>>> My personal feeling is that no
>>> actual harm is likely to be generated by this clause. A requirement to
>>> publish your code on distribution would be more onerous and arguably
>>> more dangerous, but that's not what we have here.
>>
>>This is waiting to explode into that if Trolltech, INRIA, or other
>>QPL-users publish demands for code.
>
> Why?
Imagine INRIA publishes a public request that anyone who's distributed
any changes to Ocaml's compiler to anybody send it back to INRIA. Now
this requirement has transformed into a requirement that you publish
your code on distribution, *and* it's done so retroactively, because
this license-bomb was allowed in.
> --
> Matthew Garrett | mjg59-chiark.mail.debian.legal@srcf.ucam.org
--
Brian Sniffen bts@alum.mit.edu
Reply to: