What I personally want more than a good test suite in #ActivityPub, and possibly a lower bar, is a good reference implementation.
Specifically, I'd want something that:
Implements all nontrivial requirements of the spec.
Implements a significant percentage of the data vocabulary
Can handle a nontrivial amount of traffic. It doesn't need to be optimized for it, but it can't fall over in a stiff breeze—analysis should be comparable to analysis of a prod system, just at smaller scale
I think in many ways a test suite would be more generally useful, but part of what I personally find challenging is simply determining what it even means to "handle a response."
A reference implementation would solve this.
I know of a few efforts in this regard, but most of them seem to have an iron triangle around the three criteria above:
One or two of the three only. Never all three together.
So far I have developed some heuristics when looking at implementations here.
