initial proposal for graphsync support in ipfs

[willscott]

Rendered

• [ ] Tagging @hannahhoward to tell me where i'm wrong: in particular if estimation of scope is off, and if i identified reasonable place where this plugs in & challenges.
• [ ] Tagging @aschmahmann on if there are objections to graphsync being incorporated in this way we should know about

[BigLep]

Lots of great comments and points raised by @hannahhoward . When time permits, I assume you'll respond/incorporate @willscott .

[willscott]

Yep. This definitely needs a lot more clarity. I think the product asks are still evolving quite a bit as well:

• do we want a go-IPFS node to initiate a connection to a lotus node, or vice versa?
• are we okay speaking bitswap initially?
• what sort of selector support do we need and where does that come into play (e.g. threaded up from an API command being made by a user to the go-ipfs node?)

Decisions on that path (cc @pooja ) will inform the next pass at this project scope.

[mikeal]

Can we set this PR to draft until you have time to refine it?

[willscott]

I don't see a button to do that, but no objection.

[rvagg]

@willscott over just under the reviewer list at the top right ⬈ Still in progress? Convert to Draft.