Here is a brief update from my last post. At IIW 36 on April 19, 2023, I hosted a session titled "Your DID method sucks! Change my view."
We devised a rubric for what makes a good human-centric DID method and evaluated some existing methods. You can find the notes from the session below.
Principles of a Good (Human-Focused) DID Method
-
Sufficiently decentralized (or decentralizable). This means fulfilling properties necessary for decentralization: trustlessness, no single points of failure, censorship resistance, etc. Decentralizable, meaning the design leads to decentralization even if the current conditions are not decentralized (e.g., Bitcoin is decentralizable even if only one node were running)
-
Able to scale to billions (world scale) while maintaining a low cost per transaction (< $0.50).
-
Robust feature set including create, read, update, deactivate, recovery, multisig, service endpoints, multiple key types/cryptographic agility, key delegation, etc.
-
Able to present a canonical DID state with point-in-time resolution.
-
Able to be implemented across multiple permissionless anchor systems. This allows for a diversity of implementations without mandating interoperability between chains, which could be complex and hinder other principles.
-
Limited complexity. Complexity leads to bugs; bugs lead to loss of trust. Less complex = higher diversity of implementations = more decentralization.
-
Memorable and unique names. Human readability for additional utility.
-
Flexible resolution (on/offline) and global discovery. This enables the broadest possible reach and utility of the DIDs.
Conclusions
After analyzing several common DID methods (web, ethr, key, sidetree, peer, keri, ens, cheqd) we realize that to fulfill the principles above, one needs to use either…
- A ledger-less or a multi-ledger system like KERI
- An L1 system like ETH or XTZ
- A sidechain or L2 system like a BTC spacechain
We acknowledge that none of the existing methods fulfill all principles today. The closest are the sidetree, keri, and eth/ens based methods.