Re: [Multiformats] New versions of Multibase and Multihash published

[Robin Berjon <[e-mail]>]

On 21/08/2023 15:06, Manu Sporny wrote:
> On Mon, Aug 21, 2023 at 8:39 AM Robin Berjon <[e-mail]> wrote:
>>> No, because CIDs are not file hashes, so you can't just use Multihash for them:
>>>
>>> https://docs.ipfs.tech/concepts/content-addressing/#cids-are-not-file-hashes
>>
>> I'm not sure how you reach that conclusion? You can put a CID in a URL
>> and the scheme will determine how it gets resolved. What that looks like
>> really depends on what you're trying to achieve.
> 
> All of the examples Melvin provided were specific to file hashes...
> the "mh" and "ni" schemes don't just work with any multicodec value,
> they are expected to be used with raw file hash values.

Ah, sorry, I took it as a reply to Melvin general question, not to the 
specific schemes he was exemplifying with.

> I can see the argument for "CIDs are just another type of content
> identifier that is based on a cryptographic hash -- therefore, they
> should be recognized by the mh and ni schemes", and that is /an/
> argument. It's a weak one, IMHO... a stretch, especially since CID
> v1-v3 are not identified as multihash values in the Multicodec
> registry. IPFS' own documentation is pretty clear on the matter that
> "CIDs are not file hashes".

I wouldn't use CIDs for hashes without further qualifications. The 
mapping can be sensible (e.g. for SRI, perhaps normalised around 
Iroh/Blake3, etc.) but not arbitrarily so.

-- 
Robin Berjon (he/him)
Governance & Standards at Protocol Labs
https://berjon.com/ - https://mastodon.social/@robin