rebased old BLS rework. For discussion purposes. #1233
Conversation
Rebased work from Justin.
Rebased work of Justin.
Rebased work of Justin.
protolambda
added
general:discussion
general:presentation
| ```python | ||
|
|
||
| def bls_verify(pubkey: BLSPubkey, self_signed_object: Container, domain: BLSDomain) -> bool: | ||
| return bls.bls_verify(pubkey, domain + signed_root(self_signed_object), self_signed_object.signature) |
There was a problem hiding this comment.
If we go this way (which I'm really not sold on!! I think there's a lot of value in reducing the depth of the abstraction tower that readers have to waddle through, and the verify(pub, msg, sig) format has a lot going for it given it's simple and familiar, and also symmetry with bls_verify_multiple is valuable), can we at least call it bls_verify_signed_container? Would make it clearer to readers what's going on imo.
|
I suggest we stick with current interface and go ahead with the misc cleanups (e.g. add |
JustinDrake
added
the
milestone:June 30 freeze 🥶
|
@JustinDrake You can go ahead with the misc. cleanups in the helper cleanup PR. As expressed in the closed issue, I generally think the primary part of this PR should not be considered for spec-freeze. So also dislike labeling this as "spec-freeze", for just a Instead I want to keep the discussion alive to some time after freeze, where there is less time pressure, and we can find some sweet spot (not necessarily with self-signed objects in interface, but possible) in separation of BLS and spec. |
|
Unlabeling "spec freeze", in favor of #1246 which implements Domain and DomainType, but leaves interface as is. A simple and clean solution for freeze. We can experiment more with this PR later, or not, see what the cons/pros are here. |
protolambda
removed
the
milestone:June 30 freeze 🥶
|
I'd say close this PR and revisit when we merge in the BLS standardisation post-freeze. |
Honestly, the BLS interface as-is is ok*. But if we still want to separate out application layer from the inner BLS layer, we may want to consider something alike to this PR; a rebased version of the initial work by @JustinDrake
I propose we discuss if we move forward with it, making it a post spec-freeze non-breaking change, i.e. fully backwards compatible for signatures during testnet. And keep this apart from the BLS spec-update itself, as a presentation thing.
*: ok, test-vectors will adopt the spec status to alleviate the typing inconsistency highlighted in #964