On Tue, May 20, 2003 at 04:39:30PM +0200, Henning Makholm wrote:
> Scripsit Branden Robinson <branden@debian.org>
> > On Tue, May 13, 2003 at 02:35:05PM +0200, Henning Makholm wrote:
>
> > > I wonder how the arguments I pointed to came into being, then, if I
> > > did not construct them.
>
> > Which arguments?
>
> The ones IN MY MESSAGE!
>
> > You keep saying they exist
>
> I keep giving them.
Well, it is in the message to which I am now replying that I am first
seeing most of them. But anyway...
> > > > Okay, so you and Nick Phillips will cotton to "the GNU GPL" being a
> > > > variable rather than a constant,
>
> > > Strawman.
>
> > How so? You and Nick are claiming that the "terms and conditions" of
> > the GNU GPL as applied to a work are not always the literal text of what
> > lies between "TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND
> > MODIFICATION" and "END OF TERMS AND CONDITIONS", aren't you?
>
> I say that when one constructs at cut-and-paste licence, then the
> words "this license" obviously refers to the entire cut-and-paste
> license, regardless of from where those words entered the
> cut-and-paste license.
What do you mean by "cut-and-paste"? And I think the wording of the GNU
GPL terms and conditions section makes the location of the words very
important, at least where additional restrictions are concerned.
> > Thus, the "terms and conditions" -- even in written form -- of the GNU
> > GPL as applied to a given work can vary,
>
> No they can't. For at given work they always refers to the license
> covering that given work.
...whatever that is, which is ill-defined in many cases.
> > No, that's an incidental problem. My point is that whatever a license
> > is that is the GNU GPL plus additional restrictions (consistent or not),
> > it isn't the GNU GPL.
>
> That is true. Nobody is, as far as I am aware, disagreeing with that.
Good. Then perhaps you'll agree that saying "This is licensed under the
GPL with the additional restriction that" is an invalid statement,
because such a thing is not licensed under the GPL at all.
I'd say such a statement is unacceptably misleading.
> > This is because the text of the GNU GPL is self-referential, and
> > talks about "these terms and conditions" (clause 6).
>
> If one creates a new text by using the GNU GPL, the new text becomes
> self-referential too. It does not stay GPL-referential, because what
> happens when one constructs a cut-and-paste license is *textual*
> inclusion-by-reference. One does not include the *meaning* of the GPL;
> only its *text*.
What do you mean by "cut-and-paste"? Certainly if one hacks out the
terms and conditions chunk from the GNU GPL and modifies it, one does
not have something that it textually equivalent to the GNU GPL. It's
quite obvious that one has something different.
I have to reject your assertion that "what happens when one constructs a
cut-and-paste license is *textual* inclusion-by-reference".
What happens when one constructs a cut-n-paste license is
inclusion-by-value. Whatever inclusion-by-reference is, (I have to
assume it's "This is licensed under the GPL with the additional
restriction that") it certainly cannot be inclusion-by-value. In
inclusion-by-value, the meaning of "these terms and conditions" is clear
and unambiguous. In inclusion-by-reference, it's vague, and clause 6 of
the referenced GNU GPL renders the work undistributable.
> > What the author chooses and what he manages to express are not always in
> > consistent harmony. Copyright licensees are bound first and foremost by
> > the licensor's explicit terms and conditions. Licensees who operate
> > under a "gentlemen's agreement" with the licensor in which they violate
> > the literal terms of the license do so at their own peril.
>
> Exactly. And when an author specifies that the work is goverened by a
> cut-and-paste licence, then the text of said cut-and-paste license
> *is* "the licensor's explicit terms and conditions".
What do you mean by "cut-and-paste license"?
When I think of cut-and-paste, I think of visual operations that involve
duplicating and reproducing text by means of a selection buffer. You
seem to mean a pointer, "inclusion-by-reference".
When I hear "cut-and-paste license", I think "inclusion by value". You
seem to mean "inclusion by reference".
I should note that, by far, most GPL-plus-added-restriction licenses
that we've encountered on this mailing list have been "inclusion by
reference" licenses. I.e., "This is licensed under the GPL with the
additional restriction that".
> > > On the contrary, it is obvious that if an author knowingly and
> > > explicitly licenses is work under "the conditions of the GNU GPL, with
> > > the following additional restrictions: [bla bla]", then the author
> > > intends every occurence of "THESE TERMS AND CONDITIONS" to refer to
> > > his entire combined license conditions.
>
> > You're drawing an inference. It is not obvious.
>
> I think it cannot be more obvious. It is the only way to give the
> author's utterings any sensible meaning.
Colorless green ideas sleep furiously.
Sometimes copyright holders make utterances that cannot be rendered
sensible, and/or have no meaning.
> > When an author says to me:
>
> But that is a bad example. What the author is saying to you is
>
> You can copy and distribute this work under certain conditions.
> To find out what those conditions are, take the GNU GPL and add
> to it the requirement that you can't do baz.
Nothing is being cut-and-pasted here. This is inclusion by reference.
"To see what I mean, look at the GPL", which says, "these terms and
conditions", and quite obviously means the bit between "TERMS AND
CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION" and "END OF TERMS
AND CONDITIONS" -- *in the original license text of the GNU GPL itself*.
"Add to it the requirement that you can't do baz"? This license
statement immediately falls over dead, killed by GNU GPL clause 6. We
can't distribute such a thing.
> > ..then I become confused because he is simultaneously telling me I can
> > and cannot do "baz".
>
> Yes, of course, because your'e quoting a horrible way of doing it.
It's *precisely* what you're asking me to do in my head with this
virtualized "cut-and-paste" business.
* "To find out what those conditions are, take the GNU GPL"
[sound of brain catting /usr/share/common-licenses/GPL to cognition
buffer]
* "and add to it the requirement that you can't do baz."
The result looks EXACTLY like:
Copyright 2003 Joe Blow.
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
1) You can do foo.
2) You can do bar.
3) You can do baz.
4) Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further restrictions
on the recipients' exercise of the rights granted herein. You are not
responsible for enforcing compliance by third parties to this License.
END OF TERMS AND CONDITIONS
(Okay, not *exactly* like it. Obviously the text of the GNU GPL is there
instead of my weird abbreviation. I trust I do not need to actually
physically inline a.k.a. copy-and-paste a.k.a. include-by-bvalue the
license text of the GNU GPL. I do trust you get my point nonetheless.)
> As far as I'm concerned, this discussions is still not about "is there
> a horrible way of doing this?" but "is there a valid way of doing
> it?".
Sure there's a valid way to do it.
> You can copy and distribute this work under certain conditions.
> To find out what those conditions are, take the GNU GPL and add
> to it the requirement that you can't do baz.
...is not it. It is at worst a logical contradiction, and at best
undistributable even in non-free.
> > > > Observation Three
> > > > -----------------
> > > > The FSF does not want people "forking" the GPL, and think that license
> > > > proliferation is a bad thing.
>
> > > This is true, but it does nothing to argue that a fork would be
> > > internally inconsistent.
>
> > And I have not made such an argument.
>
> Then I don't see why we are arguing at all?
I guess because you're calling
mylicense, license *char;
mylicense = license;
an example of "cut-and-paste", and not "reference"?
I don't know. You continue to bewilder me.
> > People can fork the terms and conditions of GNU GPL all they want as
> > far as I'm concerned, but if they create a license document that is
> > self-contradictory
>
> Since you "have not made such an argument" [that a cut-and-paste
> license is self-contradictory], then what's the point?
Now you seem to mean what *I* mean by cut-and-paste. No, a
cut-and-paste license, by *MY* meaning, where one duplicates the text of
the terms and conditions of the GNU GPL and then tweaks the result, is
not necessarily self-contradictory. The example you cited, which looks
like this:
> You can copy and distribute this work under certain conditions.
> To find out what those conditions are, take the GNU GPL and add
> to it the requirement that you can't do baz.
...and which is generally the genus of "GPL fork" what we encounter on
this mailing list, is emphatically *not* a cut-and-paste license. It's
inclusion-by-reference. It says, "go look over there for my terms. Now
that you've read those, add this restriction". Like I said. Bang. The
license immediately falls over dead. The licensor has simultaneously
told me I am allowed to do baz, and I am not allowed to do baz. He
needs to get his story straight, and this sort of garbage is not
acceptable in Debian. Not main, not non-free, not anywhere. It's "all
rights reserved". When a licensor cannot even make a coherent statement
of license, it is imprudent in the extreme to make assumptions about his
intentions.
> > > > However, in practice, most people don't do what the FSF recommends in
> > > > their FAQ. They just say "it's under the GPL" and slap extra
> > > > restrictions somewhere else -- perhaps in a README file.
>
> > > Which I can easily agree is internally inconsistent and may make the
> > > work undistributable.
>
> > Okay. You can easily agree -- but you're not going to???
>
> I *am* agreeing that the particular practise you describe is
> internally inconsistent. I do not agree that this is also the case for
> every *other* way of creating a (non-free) licence by adding
> restrictions to the GPL.
Where did I say that? My concern has always been with this sort of
bullshit:
> You can copy and distribute this work under certain conditions.
> To find out what those conditions are, take the GNU GPL and add
> to it the requirement that you can't do baz.
Not with people copying the terms and conditions section of the GNU GPL
wholesale and modifying it from there. They may or may not get in Dutch
with the FSF over the copyright on the GNU GPL itself, depending on
which statement from the FSF is controlling this week, but there's
nothing inherently illogical or paradoxical about such a practice.
> I'm not talking about cases where people "use the GNU GPL's terms and
> conditions as-is". I'm talking about cases where people *explicitly*
> do *not* use them as-is, but only as an expository aid for describing
> what they mean.
Including the GPL by reference as an expository aid when one is adding
restrictions is worse than useless. It elucidates nothing and, in every
instance I can think of, results a paradoxically licensed work.
Including the GPL by reference as an expository aid when one is adding
*permissions* is another story entirely. Clause 6 of the GNU GPL does
*NOT* say:
6. Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not grant any further
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
latitude with respect to the recipients' exercise of the rights
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
granted herein, or the exercise of any other rights. You are not
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
responsible for enforcing compliance by third parties to this License.
If it said this, it would cause problems for the preparation of
derivative works as discussed in clause 2.
> > > The words you used were
>
> > > || Debian interprets "this License" and "herein" to mean the
> > > || conditions of the GNU GPL expressed in its text; no more and no
> > > || less.
>
> > > which does not seem to me to acknowledge that there is a right way to
> > > add extra restrictions. (Right as in "probably legally sound", not
> > > right as in "free").
>
> > Of course there is.
>
> In that case I've misunderstood your original quote. Care to explain
> what it means, then?
If "this License" and "herein" is coming from the text of the GNU GPL,
version 2, as published by the Free Software Foundation, etc., then the
meaning seems obvious to me. The GNU GPL makes reference to itself, not
some other license document. When other people reference the GNU GPL,
they make reference to the text of the GNU GPL, version 2, as published
by the Free Software Foundation.
Other licenses that are not the GNU GPL can be self-referential, of
course. But, tautologially, they are not the GNU GPL, and cannot mean
the GNU GPL when they say, whether directly or indirectly, "this
License" or "herein". By the same token, if a license that is not the
GNU GPL *does* reference the GNU GPL, referenced text within the GNU GPL
that says "this License" or "herein" *must* mean the GNU GPL, not some
non-GNU-GPL license.
As I said before: when the GNU GPL says "this License" and "herein",
these terms are not variables. They are constants. They always and
forever will refer to the terms and conditions laid out within the same
document. It is not semantically possible for a different meaning to be
superimposed on these words without changing the content, and thus the
identity, of the document.
It boggles my mind that this requires explanation.
> > It will be if they use the GNU GPL's terms and conditions as-is.
>
> See above.
If they use the GNU GPL's terms and conditions as-is, "this License",
"herein", "these terms and conditions", and so forth refer to the GNU
GPL, not something else.
> > > But I still do not think you have offered any convincing arguments
> > > that "GPL with the following additional restrictions: [blah blah]"
> > > is a bad name to rename it to.
>
> > Well I think I have. Such a license, whatever it is, is not the GPL,
> > and will not be GPL-compatible[1].
>
> Is that an argument? Of course it is not the GPL and is not
> GPL-compatible. How does that make ia a bad idea to call it "GPL with
> the following additional restrictions: [blah blah]"? That name very
> explicitly tells the reader that the license it applies to is
> *different* from the GPL.
My turn; see above. What you describe it not useful as a license
statement and it's highly misleading as a name. Apart from that, I
think it's probably also a bad idea to give people just a name for
license (what they "call it") and expect people to construct their own
license document.
Just show 'em the freakin' license. Once that is done, the name is less
important. It could be the "PHP-Nuke! License" or the "New Improved
General Public License" or whatever.
> > > On the contrary, I think it is the honest way to go about it, rather
> > > than reproducing the entire load of legalese as "the Foobar Public
> > > License"
>
> > A license name one hasn't seen before is a damn good reason to read it
> > closely.
>
> Yes, but it is much easier to do so if the licensee does us the favor
> of exlaining how his licence differs from something we know in advance.
/me shrugs
That's certainly very nice but not mandatory. We have wdiff. We can
and should see for ourselves.
> > > and hiding the incriminating
>
> > Eh? Why are you suddenly alleging criminal activity?
>
> I'm speaking figuratively here. s/incriminating/non-free/ if you want
> to be formal.
>
> > > restrictions in the midst of vaguely-familiar licensing text that
> > > nobody is going to read closely anyway because it just looks like a
> > > revamped GPL anyway.
>
> > We here on debian-legal will read it closely. We'd better!
>
> Yes, but that does not mean that we always succeed. There has been
> several cases recently where we found that some licence we believed
> was free is really not, because we did not read it closely enough
> the first time.
>
> I don't thin we should encourage licensors to write their licenses in
> a way that will make it more difficult than necessary for us to spot
> non-free features of it.
I'm not suggesting that we do so. I am saying, however, that "the GPL
plus additional restrictions", of itself, is practically useless as a
license strategy. It is logically inconsistent, as I have endeavored to
illustrate, and for practical purposes it's highly misleading as to the
freedom of the work ("he said GPL, it must be okay") and to its
relationship with other free software ("he said GPL, it must be okay to
share the code with other GPLed works").
Let's get back to your example:
> You can copy and distribute this work under certain conditions.
> To find out what those conditions are, take the GNU GPL and add
> to it the requirement that you can't do baz.
As a matter of interpretation, it is impossible for us to, when closely
reading and thinking about the meaning of the above, to be absolutely
sure whether the licensor means to give us the right to do baz or not.
After all, the GNU GPL, to which we are referred, might explicitly give
us the right to do baz, but also says that any further restriction means
we can't distribute the work at all.
As a matter of policy, I think it is prudent to discourage licensors
from riding on the back of the GPL's success and goodwill by misleading
people as to both the license's freeness ("he said GPL, it must be
okay") and its compatibility with other GPLed works ("he said GPL, it
must be okay to share the code with other GPLed works"). They do in
fact get misled. I should have thought that the PHP Nuke license thread
made that screamingly obvious.
Maybe people shouldn't be misled, maybe they should be paranoid and
school themselves in free software licensing. But if so, what function
have we on this mailing list? If our informed opinions are to be lent
absolutely no value whatsoever, why are we wasting our time here?
I wouldn't say that people should *substitute* debian-legal's judgement
for their own, but they should permit debian-legal's judgements to
*inform* their own, to the extent that we merit respect for the
closeness of our reasoning.
> > > It is clear as day that if the author starts his source files with
>
> > > /* Copyright YYYY N.N.
> > > * This file can be copyed freely under the GNU general public license
> > > * with the added restriction that everyone who distributes it (or
> > > * derived works) must pet a cat.
> > > */
>
> > > then OBVIOUSLY he means for "THESE TERMS AND CONDITIONS" to include
> > > his additional restriction.
>
> > Except elsewhere, when speaking with his own voice, he says between the
> > words "TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION" and
> > "END OF TERMS AND CONDITIONS" that "Activities other than copying,
> > distribution and modification are not covered by this License; they are
> > outside its scope."
>
> Yes? That is not incompatible with the added restriction of the
> permission to distribute. The fact that the permission only applies to
> people who happen to pet cats does not make the permission itself
> cover other activities than distribution.
Huh? You've totally lost me here. The licensor is simultaneously
saying at least three things:
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
1) You may copy and distribute verbatim copies of the Program's source
code as you receive it, in any medium, provided that you conspicuously
and appropriately publish on each copy an appropriate copyright notice
and disclaimer of warranty; keep intact all the notices that refer to
this License and to the absence of any warranty; and give any other
recipients of the Program a copy of this License along with the
Program.
2) Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further restrictions
on the recipients' exercise of the rights granted herein.
END OF TERMS AND CONDITIONS
3) Everyone who redistributes the Program must pet a cat.
2) conflicts with 3), because the requirement in 3) is not between
"TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION" and
"END OF TERMS AND CONDITIONS".
If the licensor wants 3) to be part of his license, he cannot reference
the "GNU general public license" for his license terms and expect the
result to be distributable, since the "GNU general public license" says
nothing about "petting cats".
He needs to cut-and-paste the terms and conditions of the GNU GPL and
add the restriction mandating cat-petting to them.
> > Certainly, where that meaning can rationally be deduced from the words
> > he uses.
>
> And it can.
Not in cases where you have:
1) You can do baz, but you don't have to pet a cat when you do it.
2) When you redistribute, you're not allowed to require that people who
do baz pet a cat.
3) You have to pet a cat when you do baz.
All in the same license.
> > I don't question that they likely differ. Discerning exactly what he
> > *does* mean is another matter entirely.
>
> Exactly what he *does* mean is easy to see. There is only one way of
> interpreting the author's words. Why do you not think that is the one
> we should use?
I agree, there's only one way to interpret the author's words when he
includes the GPL by reference:
Copyright 2003 Joe Blow.
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
1) You can do foo.
2) You can do bar.
3) You can do baz.
4) Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further restrictions
on the recipients' exercise of the rights granted herein. You are not
responsible for enforcing compliance by third parties to this License.
END OF TERMS AND CONDITIONS
You can't do baz.
I can do baz, I can't do baz. It's garbage. Throw it out and ask for a
license that makes sense.
--
G. Branden Robinson | As people do better, they start
Debian GNU/Linux | voting like Republicans -- unless
branden@debian.org | they have too much education and
http://people.debian.org/~branden/ | vote Democratic. -- Karl Rove
Attachment:
pgprmjKtcye9e.pgp
Description: PGP signature