zlacker

I mean, I don't think you're incorrect, but I do think that like, semantics matters here. There is a test suite, of their implementation. Just because they have not extracted it and made it easily re-usable for alternative doesn't mean that they don't have checks to ensure regressions don't appear, you know?

They already build off of many related specifications, which have independent implementations, and a lot of the core protocol is RPC style, with schemas that they do publish. So there's already a lot of rigidity here for alternative implementations to use in a way that is extremely likely to be compliant.

I guess another way of putting it is "I don't exactly disagree with you but doing that takes work, and we're at the stage of this stuff where that work is being done, so expecting it right now feels premature to me." The spec isn't "released" or "done," it's in development.