Accepting feedback on the spec?

Hi TOSA-authors, following from the LLVM discourse, I was wondering how you would like to organize discussion around the spec (and whether there are legal processes involved to contributing)?

We have a number of feedback points around “ergonomics” of the IR that likely fall somewhere between the spec and its realization as a dialect in MLIR (i.e. when to use operands vs attributes, availability of explicit broadcasting ops, some questions about the possibility of expanding the lookup table facility, etc). Before just inviting the people on my team in who have that feedback, I wanted to check point on the process a bit. Should we just have such discussions here and then if items only impact the MLIR implementation, move them to that venue?

We absolutely want to have these discussions as soon as possible and are very keen that TOSA is developed in the open and in a collaborative way, with contributions from many others. However, we also want to make sure that TOSA remains a royalty-free standard which anyone can implement. In order to achieve that second goal, we need to make sure all the contributions we accept into the specification are provided with appropriate rights to any IP owned on the contribution. We’re currently working out some form of contribution agreement to enable us to do that, which might take a few weeks. In the meantime, I can at least write up the contribution process on the wiki somewhere. Once I have that, I’ll post a link on this thread.