search

decentralized-identity / peer-did-method-spec

A rich DID method that has no blockchain dependencies. The verifiable data registry is a synchronization protocol between peers.
https://decentralized-identity.github.io/peer-did-method-spec/index.html
Apache License 2.0
27 stars 17 forks source link

switch delta format #24

Open dhh1128 opened 3 years ago

dhh1128 commented 3 years ago

@OR13 suggested here that we switch the format of deltas to use signed-ietf-json-patch... https://github.com/decentralized-identity/did-spec-extensions/blob/master/parameters/signed-ietf-json-patch.md.

I think this is probably a great improvement. I wanted to ask a couple clarification questions first.

  1. It's important to peer DID CRDT update logic that deltas can be applied in different orders. They have some internal properties that describe their preconditions, but if A->B->C produces the same state as A->C->B, and the intermediate difference is not one that would alter the outcome, either sequence should be allowed. From my reading of signed-ietf-json-patch, it looks like this is not a problem. However, I would be concerned if the patch requires a specific pre-existing state such that it forces absolute ordering. I would also be concerned if the patch doesn't allow any statements about pre-existing state to force approximate ordering. Does anybody know if we can cross this off my worry list?

  2. What dependencies do we incur by using this patch mechanism?

OR13 commented 3 years ago

Another thing to consider is the ION / Sidetree is no longer even using initial-state... because they have opted to make the identifier compound.... AFAIK did:peer could do the same and avoid both initial-state and signed-ietf-json-patch.

Here is the difference:

Option 1

did:example:123?signed-ietf-json-patch=JWS

Option 2

did:example:123:JWS

The DID Method then decided how to handle the "long" and "short" form DIDs.

llorllale commented 3 years ago

Where are we on this?

dhh1128 commented 3 years ago

The short answer is that nothing has happened on this because we've been focused elsewhere.

I am not opposed to a change in concept, as long as the necessary characteristics of the delta don't change. However, I think this will continue to be back-burner for a while.

OR13 commented 3 years ago

I chatted with some other did peer implementers, and concluded that is it better to just use did:key, until KERI based methods catch up and can replace did peer. If you wondering if you should spend time on did peer, my 2 cents is to help KERI, use did:key or use sidetree.