Kevin Atkinson

> So I think the priority is to first decide at the specification level what role does the prefix play so the user has a clear mental model of it,...

If part of the confusion is that the `Prefix` structure has a `Bytes()` method I think we should just remove it as I am not sure how much utility it...

And maybe we should just eliminate the use of the word `Prefix` for this. I like using `CidInfo` for a struct that is just about the information in a Cid...

@Stebalien that may be true, but now that we can put strings in maps how will `map[cid.Cid]string` serialize since the key is a JSON object and not a string?

Right my point is with the current serialization of a CID a `map[cid.Cid]string` won't serialize to valid JSON. This could be a problem..... EDIT: I didn't actually try this so...

So after testing and a little more research it seams `map[cid.Cid]string` won't serialize anyway. To fix this we would need to implement the `encoding.TextMarshaler` and `encoding.TextUnmarshaler` interface which just convert...

> Note: we'll have to be very careful with this. This is the kind of optimization that could end up hurting performance (or even memory usage) more than it helps...

It is important to note that the 64 bytes comes from the maximum size modern crypro. hashes output (512 bits). If we set the limit lower than this will we...

> this should be optional/experimental behaviour. Otherwise, pre-existing tests fail because the CID is different Why?

@richardschneider the automatic use of identity hashes will require a command line flags that is not enabled by default. See https://github.com/ipfs/go-ipfs/pull/4910 for a proof-of-concept implementation.