PeerId string identifier

Type:

Question

Description:

Currently, js-libp2p uses PeerId objects to identify peers, which are compared using equals method and printed with toB58String. In many cases, the b58 string is used to index peers (eg: in Maps, Objects, Sets, etc) and there, implicitly used for id equality.

Has any thought been given to using b58 string encoded peer-ids as the canonical peer identifier throught the codebase? In this case, the keystore would be more heavily relied on to retrieve public/private keys since they wouldn't be (optionally) attached as in PeerId objects. In many cases, checking the validity of a string identifier would still be required, but strings are convenient to use for indexing and equality checking. Perhaps this may also result in general speedups and reduced memory usage, as string equality checking is generally faster than Buffer equality checking and public keys can all be stored in one place.

I definitely haven't thought thru the intricacies, just more an idea after seeing lots of id.toB58String() in my own code, and seeing that peer.ID is also a string in go-libp2p.

Curious more than anything if this is being considered or on the roadmap, and if this has been discussed before.

Start by reviewing PeerId usage, calls to toB58String, and how the keystore supplies keys. Compare string identifiers with PeerId objects for equality, indexing, validation, memory, and performance; done means documenting a clear decision about whether canonical b58 identifiers should be adopted or deferred.