How about a pragmatic "solution":
- NVidia will provide documentation to their hardware (similar to what AMD does) and,
thus, finally shows real support of Linux (and OSS). Additionally, they will support the OSS driver developers.
- In return the use of the kernel API by the proprietary driver will be tolerated.
I know this does not solve the legal issue, hence, let's stick to "pragmatic" solution.
But *I* think this would really solve quite a bit:
- Users of the blob see Optimus support and more (Project Denver related dma-buf stuff , etc.).
At the same time, they should know this tainted configuration is completely unsupported and even represents
an enhanced security risk.
- Users that care about OSS drivers and the GPL will significantly benefit from the documentation and enjoy
support from the kernel developers.
Yes, it sounds like horse-trading and it's full of "goodwill".
But it's exactly what's missing from NVidia until now.