Determine the Poretitioner license

[thequicksort]

Description

This task is to figure out our license for Poretitioner.

While we have a number of options, in general we're bounded by the strictest license of our dependencies. This will largely depend on what proprietary packages we end up using, which is why I suggest we save this step til towards the end of the project (or sooner if we're confident we have all the tools we need). The only dependency I've found that has a proprietary dependency is Torch/TorchVision
(pytorch depends on the Intel Math Kernel (which is Intel Simplified Software License (ISSL))

Pytorch and ISSL

While it's technically not free software, the good news is that according to the ISSL license FAQ:

• Redistribution is allowed
• There are no restrictions on redistribution
• Yes, you can include Intel MKL in a public container. (e.g. Docker)
  It seems like they're mostly just about people reverse engineering the software or using the Intel name, so as long as we don't do those two things, we should be okay.

I'd still recommend that we talk to at least one other lab working in computer vision or machine learning, just to be sure.

Nix and packages

By default, Nix won't build your package if you declare a dependency on a package that is, or depends on, non-free software.

This can be modified by either using the

config.nix

Nix config can be configured to allow unfree packages: https://nixos.org/nixpkgs/manual/#sec-allow-unfree

e.g. through confix.nix

 config = {
     allowUnfreePredicate = pkg: builtins.elem (lib.getName pkg) [
      "pytorch"
      "pytorchWithoutCuda"
    ];
  };