Rework go-ipfs to use content IDs version 0

[whyrusleeping]

This is a refactor to replace all object identifiers code with content ID version 0. This is a drop in replacement for the existing key.Key and will not change any outward behaviour. This is just a preparatory refactor for the switchover to CIDv1 and ipld

[Kubuxu]

This is resolved in #3162

[Kubuxu]

I went over it, can't find anything wrong with it at first sight but I would love if @lgierth went over it too.

(also it is in conflict with master now, probably due to changes in the blockstore code I pointed out).

[whyrusleeping]

One change i could probably do separately is the Object abstraction in Blockservice. It doesnt need to be done here, but it does simplify things a bit I think. If anyone thinks that it should be done separately I will go ahead and do that.

Other than that, I think that the pin code changes should be looked over closely. @lgierth if you could pay special attention there that would be great :)

[ghost]

This is a refactor to replace all object identifiers code with content ID version 0.

Is the following accurate? It doesn't replace all usage of key.Key -- from the block layer on inwards we assume raw data (where s/Data/RawData). The layer above that will be the one dealing with the different semantics of various CIDs (soon).

I'm looking over git diff pin/ with this assumption in mind, mh, tricky.

---

// an Object is simply a typed block
type Object interface {
    Cid() *cid.Cid
    blocks.Block
}

👍

[whyrusleeping]

@lgierth yeah, youre right. the key.Key is still around at the block layer. We're talking about making the CIDs go all the way down to disk, it has its pros and cons, but it would drastically increase the complexity of this change, and is definitely something we can do in a follow up PR.

[ghost]

I assume this is explicitly v0 because we don't wanna start using v1 right now already

[whyrusleeping]

Yeah, but thinking about it, we could change that to just be a cid.Cast and it would work just fine

[ghost]

Nah it's fine, one silly mistake and we'll have to maintain it :D:D

[ghost]

LGTM 👍

• pin/ could see the functions named after Key (e.g. Pinner.DirectKeys()) renamed
• some of pin/ still uses Key, which is mostly the GC code -- I can understand why you didn't wanna touch it right now, it can get messy and we can take care of it later.
• these two made it a bit confusing at first, but I think it's good to go

[ghost]

And I guess this PR has the nice side-effect of reducing the overall dependency on the blocks package :)