llvm dev - Nov 2016 - RFC #2: Improving license & patent issues in the LLVM community

On Thu, Nov 3, 2016 at 9:13 AM, David Chisnall <David.Chisnall at
cl.cam.ac.uk>
wrote:
> On 3 Nov 2016, at 15:55, Daniel Berlin <dberlin at dberlin.org>
wrote:
> >
> >> I and the other libc++ contributors were all happy to have our
code
> relicensed under the MIT license (or contribute it under those terms
> originally), so there is some pretty clear evidence that we explicitly did
> *not* require this attribution.
> >
> > That license does, in fact, require attribution:
> > "The above copyright notice and this permission notice shall be
included
> in all copies or substantial portions of the Software. “
>
> Again, unless I’m misreading this, there is no requirement that code that
> merely uses libc++ headers include a libc++ attribution.

You are misreading this :)

I can tell you it's not even a consensus, i have not found a single lawyer
who believes otherwise.

I can forward you the discussion that occurred around this if you like (i
have to acquire permission from the folks involved first). There was 100%
universal agreement on this.

> That was the explicit requirement for the switch to the MIT license.  The
> proposed new license would mean that this is only true for code compiled
> with LLVM.
>
> We must include the copyright and attribution when we distribute FreeBSD,
> because we include the whole of libc++.  We do not need to separately
> include it in programs that simply link to libc++ (irrespective of how they
> were compiled) in packages, because they do not contain ‘substantial
> portions’ of the code (probably - it’s a bit of a grey area with respect to
> template-heavy C++).
>
This is not correct ;)

>
> >> Indeed, we relicensed under the MIT license (after long
discussion)
> specifically to avoid this requirement.
> >
> > If we did, we screwed it up then. We probably meant the zlib license,
> which does not require attribution  ;)
>
> Assuming that your interpretation of this clause - which is the first time
> I have ever heard it interpreted to include either linking or including
> headers - is correct.

This is interesting, as i've said, i've literally never heard it
interpreted otherwise.

> I’d be quite surprised if it were, as the lawyers that reviewed it at the
> time were happy that it did not impose these constraints, but it’s
possible.
>
Like I said, there was discussion of it on the open source counsel mailing
lists i belong to, and there was universal agreement that it required
attribution.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161103/fb0c9cd9/attachment.html>

On 3 Nov 2016, at 17:11, Daniel Berlin <dberlin at dberlin.org>
wrote:
> On Thu, Nov 3, 2016 at 9:13 AM, David Chisnall <David.Chisnall at
cl.cam.ac.uk> wrote:
> On 3 Nov 2016, at 15:55, Daniel Berlin <dberlin at dberlin.org>
wrote:
> >
> >> I and the other libc++ contributors were all happy to have our
code relicensed under the MIT license (or contribute it under those terms
originally), so there is some pretty clear evidence that we explicitly did *not*
require this attribution.
> >
> > That license does, in fact, require attribution:
> > "The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software. “
> 
> Again, unless I’m misreading this, there is no requirement that code that
merely uses libc++ headers include a libc++ attribution.
> 
> You are misreading this :)
> 
> I can tell you it's not even a consensus, i have not found a single
lawyer who believes otherwise.
> 
> I can forward you the discussion that occurred around this if you like (i
have to acquire permission from the folks involved first). There was 100%
universal agreement on this.
>  
> That was the explicit requirement for the switch to the MIT license.  The
proposed new license would mean that this is only true for code compiled with
LLVM.
> 
> We must include the copyright and attribution when we distribute FreeBSD,
because we include the whole of libc++.  We do not need to separately include it
in programs that simply link to libc++ (irrespective of how they were compiled)
in packages, because they do not contain ‘substantial portions’ of the code
(probably - it’s a bit of a grey area with respect to template-heavy C++).
> 
> This is not correct ;)
>  
> 
> >> Indeed, we relicensed under the MIT license (after long
discussion) specifically to avoid this requirement.
> >
> > If we did, we screwed it up then. We probably meant the zlib license,
which does not require attribution  ;)
> 
> Assuming that your interpretation of this clause - which is the first time
I have ever heard it interpreted to include either linking or including headers
- is correct.
> 
> This is interesting, as i've said, i've literally never heard it
interpreted otherwise.
>  
> I’d be quite surprised if it were, as the lawyers that reviewed it at the
time were happy that it did not impose these constraints, but it’s possible.
> 
> Like I said, there was discussion of it on the open source counsel mailing
lists i belong to, and there was universal agreement that it required
attribution.
> 
If your interpretation is correct, then almost every graphical X11 application
is infringing the license with respect to libX, the authors of the license were
incorrect as to what it does, every project that I’ve been involved with that
has used it has adopted it assuming that it means something different, and every
lawyer that I’ve spoken to at companies adopting such code has been wrong.  This
may be the case, but you’ll forgive me if I take your opinion with a grain of
salt here.

The LLVM project's intention on moving libc++ and compiler-rt to the MIT
license was that no code that simply linked to unmodified versions of these
libraries would be required to contain attribution.  A quick look on
opensource.org points to TL;DRLegal, which concurs with this: you must include
the copyright notice in a copy of the library, but you are not required to
include any attribution.  In fact, a quick skim of similar sites includes the
same consensus, so I’m not sure who is on this mailing list that you allude to,
but they seem to have some quite different opinions to the normal consensus.

David

On Thu, Nov 3, 2016 at 2:04 PM, David Chisnall <David.Chisnall at
cl.cam.ac.uk>
wrote:
> On 3 Nov 2016, at 17:11, Daniel Berlin <dberlin at dberlin.org>
wrote:
> > On Thu, Nov 3, 2016 at 9:13 AM, David Chisnall <
> David.Chisnall at cl.cam.ac.uk> wrote:
> > On 3 Nov 2016, at 15:55, Daniel Berlin <dberlin at dberlin.org>
wrote:
> > >
> > >> I and the other libc++ contributors were all happy to have
our code
> relicensed under the MIT license (or contribute it under those terms
> originally), so there is some pretty clear evidence that we explicitly did
> *not* require this attribution.
> > >
> > > That license does, in fact, require attribution:
> > > "The above copyright notice and this permission notice shall
be
> included in all copies or substantial portions of the Software. “
> >
> > Again, unless I’m misreading this, there is no requirement that code
> that merely uses libc++ headers include a libc++ attribution.
> >
> > You are misreading this :)
> >
> > I can tell you it's not even a consensus, i have not found a
single
> lawyer who believes otherwise.
> >
> > I can forward you the discussion that occurred around this if you like
> (i have to acquire permission from the folks involved first). There was
> 100% universal agreement on this.
> >
> > That was the explicit requirement for the switch to the MIT license.
> The proposed new license would mean that this is only true for code
> compiled with LLVM.
> >
> > We must include the copyright and attribution when we distribute
> FreeBSD, because we include the whole of libc++.  We do not need to
> separately include it in programs that simply link to libc++ (irrespective
> of how they were compiled) in packages, because they do not contain
> ‘substantial portions’ of the code (probably - it’s a bit of a grey area
> with respect to template-heavy C++).
> >
> > This is not correct ;)
> >
> >
> > >> Indeed, we relicensed under the MIT license (after long
discussion)
> specifically to avoid this requirement.
> > >
> > > If we did, we screwed it up then. We probably meant the zlib
license,
> which does not require attribution  ;)
> >
> > Assuming that your interpretation of this clause - which is the first
> time I have ever heard it interpreted to include either linking or
> including headers - is correct.
> >
> > This is interesting, as i've said, i've literally never heard
it
> interpreted otherwise.
> >
> > I’d be quite surprised if it were, as the lawyers that reviewed it at
> the time were happy that it did not impose these constraints, but it’s
> possible.
> >
> > Like I said, there was discussion of it on the open source counsel
> mailing lists i belong to, and there was universal agreement that it
> required attribution.
> >
>
> If your interpretation is correct, then almost every graphical X11
> application is infringing the license with respect to libX, the authors of
> the license were incorrect as to what it does, every project that I’ve been
> involved with that has used it has adopted it assuming that it means
> something different, and every lawyer that I’ve spoken to at companies
> adopting such code has been wrong.

I can't speak to this, at all. My guess is you and they have a different
idea of what the words you were using mean.

This may be the case, but you’ll forgive me if I take your opinion with
a
> grain of salt here.
>
You are of course, welcome. I can only tell you what i know.


>
> The LLVM project's intention on moving libc++ and compiler-rt to the
MIT
> license was that no code that simply linked to unmodified versions of these
> libraries would be required to contain attribution.  A quick look on
> opensource.org points to TL;DRLegal, which concurs with this: you must
> include the copyright notice in a copy of the library, but you are not
> required to include any attribution.

Uh, what?
It says exactly what i said:
https://tldrlegal.com/license/mit-license

"You must include the copyright notice in all copies or substantial uses of
the work."
and
"You must include the license notice in all copies or substantial uses of
the work."

That last part is the attribution requirement.
It very clearly says you must include it in all copies or use of the work.

The work includes the header and libraries, and thus, all copies or uses of
the work includes things that include things from the headers, or link code
into the libraries.
For MIT, the attribution *is* the copyright and license notice.

In fact, for every different copyright notice in a piece of MIT licensed
software, you must reproduce *that copyright notice* and copy of the
license.



In fact, a quick skim of similar sites includes the same consensus,


 Yes, they all agree that you need to provide the attribution, see above.
So what exactly are you trying to argue here?


so I’m not sure who is on this mailing list that you allude to, but
they
> seem to have some quite different opinions to the normal consensus.
>

I think you pretty much just proved yourself wrong here?

Which means my guess is you think attribution means something else.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161103/6fa23126/attachment.html>