feat!: split crates into multiple to isolate breaking changes

[thomaseizinger]

Resolves #267.

[vmx]

I was the one saying "why not keeping the serde-codec feature". Seeing the code, I don't it makes sense to have the serde and serde-codec feature. As this will be a breaking release, I think just going for serde is fine, it should be easy to change downstream.

[thomaseizinger]

I was the one saying "why not keeping the serde-codec feature". Seeing the code, I don't it makes sense to have the serde and serde-codec feature. As this will be a breaking release, I think just going for serde is fine, it should be easy to change downstream.

I am fine either way, I don't think it is too bad keeping it for now and mentioning it in the changelog that it is deprecated. Happy to remove it as well.

[thomaseizinger]

As this will be a breaking release

My take on this would be that we should still try and make upgrading as easy as possible.

[thomaseizinger]

Thanks for the prompt review @vmx! I addressed/responded all points, please take another look :)

[vmx]

I think we are getting close!

[thomaseizinger]

@vmx Is something blocking this from being merged?

[thomaseizinger]

Given the size of this change, I would suggest that we first cut an RC of everything and ask people if they are happy with this interface.

[vmx]

Thanks a lot for all the work, nothing is blocking a merge.

Before doing a release, I'd like to get other breaking changes in, like removing the identity hash. Also some documentation (which I guess should go into the changelog) should be added to explain the upgrade path to the new structure.

I hope that downstream dependencies like libp2p, FVM, Forest etc would try things out to see if they can smoothly upgrade to this new version.

[thomaseizinger]

Before doing a release, I'd like to get other breaking changes in, like removing the identity hash. Also some documentation (which I guess should go into the changelog) should be added to explain the upgrade path to the new structure.

Definitely in favor of this. Happy to help, just ping me on some issues on what you would like to get done.

[thomaseizinger]

I would still very much in favor of cutting an RC first and potentially an RC of multiaddr where we depend on the RC of multihash, just so downstream users can give us some feedback before we commit to the new interface.