Eric Myhre

We could consider use of [printer.Sprint](https://pkg.go.dev/github.com/ipld/[email protected]/printer#Sprint) for this now, which was [introduced since this issue was first raised](https://github.com/ipld/go-ipld-prime/pull/238/). It might be overkill for this case (?), but might be the...

Not to take the fire away, but real quick: the "Union typekind acts like Map" one is actually clear and true. I was about to remove that particular comment :)...

+1 to probably favoring Option 2. (Possibly calling it `Kind_Contextual`?) I might also offer an Option 4: that we introduce another enum for this concept of "kinds that a typekind...

Hm. One _non_ issue: allocation amortization. It's a noteworthy issue where this pattern appears in the gogen code generator and in the basicnode implementation, but I don't think it applies...

Hm. Sounds potentially legit to me. I'd like to spend a moment wondering out loud if it will have any effect on maintainability, and performance, feature range, and the library's...

We may want to leave some function stubs even beyond a deprecation period, which redirect to the bindnode behaviors? It may help keep guiding people to the right place if...

I think the main distinction is, while with absent values, one may write: ``` node, err := parent.LookupByString("foo") if err != nil { handle(err) // handle other errors } if...

I agree, returning both a valid value and an error response is not very idiomatic. I'll strike that suggestion from the original post.

> This seems like the most controversial part of this to me; it seems like you're proposing a loss of information for typed nodes. [...] I assume this is to...

This is roughly the same ask as https://github.com/ipld/go-ipld-prime/issues/213, no?