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

>
>
> > For clarification: Is my interpretation incorrect?  If I compile code
> with GCC, which uses templates from libc++ headers and therefore results in
> libc++ code being inserted into the resulting binary, am I required to
> abide by clause 4 of the Apache license and include the libc++ attribution?
> >
> > Yes.
> > But, AFAIK, this is deliberate.  IE the view is that in this case, you
> *should* be giving attribution.
> > So this is at least "not a bug", regardless of whether
it's liked or not.
>
> I believe that this would be a show-stopper for FreeBSD’s use of libc++
> and compiler-rt.  As such, I would strongly oppose this and would not
> consent to any of my code in libc++ being relicensed under the proposed
> terms.  This would require that we audit all 30K packages, find the ones
> that will (when we ship binaries) link libc++, and add the attribution in
> the right places.
>
> Whose view is ’the view'?

The consensus i saw here and elsewhere. I'm happy to be shown to be wrong.

> 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. "

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  ;)

> This seems like a step that is directly contrary to the intentions of the
> original authors.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161103/3af0a355/attachment.html>

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.  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++).
>> 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.  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.

David

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>