Make CID versions multicodecs

[Stebalien]

This way, we can trivially disambiguate CIDs from, e.g., raw multihashes.

(To do this, we'd need to steal 0x1 from unary for CIDv1).

[Stebalien]

@diasdavid and I discussed this offline but I figured it should be posted here to keep things public.

We and came across the theoretical problem of handling non-multibase CIDv18 (18 = 0x12 = SHA2-256 multicodec) and decided that we could just skip that or define CIDv0 = CIDv18.

[hobofan]

Is there a way to get this moving? Are there any blockers that need to be discussed?

I'm currently in the process of drafting our new format for rlay-project/rlay-ontology, which strongly revolves around CIDs.
In our current (older) format, we always have to have some knowledge which fields of a struct have which types (CID vs multicodec), and for the fields that accept both, we have a not so pretty "try to parse as CID and if that fails it's a multicodec" logic. Adding CIDs as proper multicodecs would be great step towards simplifying this.

[rvagg]

I think the next step to move this forward is to open a PR against table.csv to add something like cid, ipld, 0x1, CIDv1, you're welcome to do that @hobofan and maybe we can move forward. The step after this might be to implement the spec change detailed @ multiformats/cid#25 to cement it.

0x1 isn't in here anymore, I believe it was extracted when multibase was split out and along the way somewhere base1 / unary was dropped entirely, so it's up for grabs here without needing to do conflict resolution.

I'm still not clear on why #94 is necessarily a different thing though tbh. It seems to be part of the same bundle of concerns and just a matter of definitions.

[Stebalien]

#94 is proposing adding a new multicodec for CIDs. So a multicodec-identified CID would be <cid-multicodec><v1><ipld-codec>.... This proposal is to just treat v1 as a multicodec (<v1><ipld-codec> -> <cid-multicodec><ipld-codec>...).

[oed]

Hey @Stebalien @rvagg I'm happy to create the PRs needed to help move this forward, as I'm interested in using multicodec similarly to CIDs without it causing conflicts. Is there anything I need to be aware of in terms of the two different options? (to me your approach @Stebalien makes the most sense)

[Stebalien]

I believe all we need is two PRs:

1. A PR to the multicodec (this) repo reserving 0x1 for cidv1, ipld, 0x1, CIDv1.
2. A PR to the CID repo (https://github.com/multiformats/cid) replacing cid-version and version in the "cidv1" spec with `multicodec-cidv1".