Then there is no concrete protocol, because everyone does different things. Nostr is in need of standardization to scale further, imho.

I'm building something using python, and I'm running into trouble with python libs for nostr so much that I'm considering just using nak as a system library and making all my calls to it. Seems fucky so I've been putting off work on this particular project while I overanalyze. Would you recommend such a design choice?

The point is that there is no need for a single implementation in order for current apps to interoperate. They are doing great even not sharing any codebase.

Just build from scratch. It's not that hard.

I see the opposite problem. When I first discovered nostr, about a year after the protocol was initially specified, I imagined it as a general message transport protocol. Then, fiatjaf said, and I may be paraphrasing, "nostr is not a general message transport protocol". So I decided okay, where is this going. I like the living network, organic evolution, minds discovering possibilities and making things real. But I also like constraints, a bucket to ultimately fill, a goal, "it works now, have fun." What I see now is everyone trying to fit their use case into nostr, and an explosion of NIPs, a situation that looks very http-esque. There are nips for single client use cases, and then generalized nips for general use cases so ambiguous that they will confuse interoperable clients. Now we are even at the point that we have different canonical registries of approved nips and permissionless nip implementation. It looks like not a protocol at all, just a guideline for what ultimately may be an unmaintainable mess, ripe for conquest by those who seek to control information distribution. I think a good balance of "this is what it is for" and "do what you want with it" is desirable, and I'm seeing that, but in an unstable equilibrium, and I fully expect to be disappointed by the future with regard to this thing we have.

It's in the public domain. It's probably partially plagiarized by many even if not imported as a library.