Andrew Twyman

This is an awkward workaround, but could you avoid this with a custom TypeWriter which ignored its primary Type and just took all of its type parameters from tag values?...

I think that null optional constants were intentionally never implemented because they seemed meaningless (why do you need a constant for null). Here's a comment in the test suite which...

@gibutm I was the reviewer on the initial Python prototype in Djinni, but it was enough years ago that I can't remember full details. I can suggest where you might...

Looking at this literally, I don't think it makes sense for all of the functions. Constructing the class is costly, and requires a private key, so shouldn't be required to...

That seems fine. The hash is the most expensive thing to be sure gets reused, but I think reusing the secret scalar skips some more computations, and is already a...

Ah, so it does. Sorry, I missed that in my scan of the code. I don't really understand the details of how the secret scalar is used in circuits (as...

> There's another paragraph for signing messages: https://datatracker.ietf.org/doc/html/rfc8032#section-5.1.6. I didn't check it though. That's what I skimmed, and I see 3 different SHA-512 hashes. Even if blake is replacing that,...

I think you're welcome to prioritize this however you wish. It would be a performance improvement specifically to the use case of signing many messages with the same key. I...

For non-breaking changes, my primary concerns to understand would be the impact on performance and bundle download size. circomlibjs has a very large set of poseidon constants for all sizes,...

Check if I'm reading your performance chart right: It seems that the mean is much slower than the other libaries, but the median is much faster. If that's accurate it...