gs1.1 direct connect

[wemeetagain]

Implements https://github.com/libp2p/specs/blob/master/pubsub/gossipsub/gossipsub-v1.1.md#explicit-peering-agreements

Follows the go-libp2p-pubsub implementation closely

[vasco-santos]

All js-libp2p interfaces collected into ts/interfaces.ts (This should likely be refactored away into a separate library eventually, eg: into js-libp2p directly or chainsafe/libp2p-ts. But for now, its extremely convenient to tweak the interfaces as needed all in one place.)

I am ok on having this for now. But I would like to have us to generate the types declarations via js docs with aegir

gossipsub constructor signature changed significantly, now (libp2p: Libp2p, options: object)
This will require a significant rewrite of all existing tests, at least the sections that create new gossipsub objects. I've begun to do this, adding libp2p, libp2p-websockets, libp2p-noise, libp2p-mplex as dev dependencies. Unfortunately, I've run into issues with browser compatability, so just decided to push what I have for now. Hopefully we can discuss and figure out how to proceed.

You will need to have a signal server for the tests. Check it out this example:

• https://github.com/libp2p/js-libp2p/blob/master/.aegir.js - this runs a node peer that browsers will interact with

This example allows us to test browser considering that a browser node will connect to it directly. I would recommend that we setup here this kind of server but with the server being a circuit relay. We would spin up n nodes, that would listen on the circuit relay peer for new connections from the other peers. Then, we can have configuration files for browser and node like here: https://github.com/libp2p/js-libp2p/tree/master/test/utils , but I would also have a function there to compute the listen multiaddrs to use. That is, we do not need to connect via the relay in node, just when running browser tests.
What do you think?

[vasco-santos]

Maybe a new PR just adding libp2p as a parameter to have the tests good before getting this in?

[vasco-santos]

When we refactor pubsub, I think we should also send libp2p to it. What do you think?

[wemeetagain]

for sure 👍

[wemeetagain]

Maybe a new PR just adding libp2p as a parameter to have the tests good before getting this in?

Yeah, I'll split this up into two PRs:

first collecting interfaces and adding libp2p as a parameter, including passing tests,
second adding direct peer connections

[vasco-santos]

@wemeetagain thanks 👍

For the tests, you can get ideas from my WIP work for the rendezvous:
server: https://github.com/libp2p/js-libp2p-rendezvous/blob/feat/rendezvous-protocol-full-implementation/.aegir.js
create node util: https://github.com/libp2p/js-libp2p-rendezvous/blob/feat/rendezvous-protocol-full-implementation/test/utils.js#L38
connect test: https://github.com/libp2p/js-libp2p-rendezvous/blob/feat/rendezvous-protocol-full-implementation/test/rendezvous.spec.js#L70

It is working over circuit-relay for connecting the nodes

[wemeetagain]

Ok, awesome, I'll give it all a shot. Thanks for the examples, super helpful.

[codecov-commenter]

Codecov Report

Merging #99 into gsv1.1 will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           gsv1.1      #99   +/-   ##
=======================================
  Coverage   91.66%   91.66%           
=======================================
  Files           2        2           
  Lines          24       24           
=======================================
  Hits           22       22           
  Misses          2        2           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b040dc8...7103b83. Read the comment docs.

[vasco-santos]

LGTM!