[llvm-dev] RFC: Improving license & patent issues in the LLVM community

Joerg Sonnenberger via llvm-dev llvm-dev at lists.llvm.org
Thu Oct 22 21:54:58 PDT 2015

On Wed, Oct 21, 2015 at 09:10:40AM -0700, Daniel Berlin wrote:
> Let's start with: In just about every country in the world, anyone
> contributing on behalf of their company are exercising their employers
> copyright (in most cases, even if they do it in their "spare time", since
> most people misunderstand what the law grants them there).

At least under German law, there is a clear separation between what I do
during work time and outside. Ignoring questions like anti competition
issues, my employer has no claims to what I do outside work time. That
leaves the part of "contributing on behalf of their company".

> Both actual (They told me i could contribute), agent (I am also actually
> authorized to contribute anyway), and apparent authority (Everyone else in
> the community would normally believe i have authority to contribute, and
> thus, there is apparent authority, regardless of whether Google said i
> could contribute) would all bind Google when i contribute stuff.

I don't think you will be able to convince a German court of that. I'm not
a lawyer, so I can only base this on my experience conducting business
on behalf of my company. *Any* contract with binding and (practically)
irrevocable conditions so far required written proof of authorisation.
Purchasing physical goods is normally exempt, because it is revocable.
Prepayment practically removes the binding conditions (to pay) and
that's how many smaller immaterial goods are bought. But the general
opinion is that apparent authority to act on behalf of the company is
very limited. I'm not talking about falsifying proof of authority, but
having to show it in first place.

For the specific situation of contributions to LLVM, I believe two
different cases apply:
(1) The contribution is considered an unlicensed copy OR a confidental
data. I don't know if it is even possible to stop third parties from
using such material after publication, but it can certainly be handled
as "planned" publication with the company claiming damages from the
contributor. In practise, I would ask (as project) for proof of authority
for significant patches from contributors working for a company competing
in this area, but that's about it. That's only because the hassle in
case of problems is too big.
(2) The contribution touches on company patents. As written above,
making assumptions about non-trivial financial contracts is pretty much
invalid. I would certainly *expect* that noone below medium management
or the company law depatment is authorised to sign such a contract,
directly or indirectly. That's why I am a bit speechless that it would
be considered reasonable in the USA.


More information about the llvm-dev mailing list