Document /http

[MarcoPolo]

This is meant to clarify what libp2p is expecting of the /http component and avoid folks creating conflicting interpretations of the component. This purposely does not include anything libp2p+http (but hints at it) with the goal of clarifying expectations of the /http component early.

Please take a look:
@lidel
@marten-seemann
@willscott
@masih

This was prompted by ipni/go-libipni#42 + historical issues referenced in ipni/go-libipni#42 (comment)

[thomaseizinger]

Currently, we expect to immediately run multistream-select after a connection is established. Connections are described by multiaddr.

How does this interact with the /http transport component?

I am guessing we don't run multistream-select on connections using the /http component?

[thomaseizinger]

Currently, we expect to immediately run multistream-select after a connection is established.

Actually, that is not true for QUIC.

Anyway, it might be worth mentioning this briefly.

[MarcoPolo]

We know this is an HTTP transport. Multistream-select is not used here. The specifics I believe belong in the libp2p+http doc since it's how libp2p uses the HTTP transport.

[BigLep]

Thanks for creating and getting this written down.

[mxinden]

With my limited knowledge of the libp2p+http effort, this looks good to me apart from the smaller comments below.

[BigLep]

Related issue: multiformats/multiaddr#63

[BigLep]

This topic came up during the 2023-06-01 IPFS Implementers Working Group (recording should go live by end of week). Everyone was told to engage here. We expect to use the 2023-06-15 occurrence to close out on this (i.e., resolve anything that wasn't handled async). Anyone is welcome to join that call (details).

[BigLep]

@MarcoPolo : I know you want to have more sync time during https://pl-strflt.notion.site/IPFS-Implementers-Sync-2023-06-15-b8556243125442f2b0bf48c5db90d056?pvs=4 . I assume you'll consolidate down the open questions and areas of disagreeement in advance.

[thomaseizinger]

One question regarding the semantics of TLS, other than that, this is a really good document and I am looking forward to this!

[MarcoPolo]

All outstanding concerns have been addressed. We brought this up in the ipfs implementers meeting and no objections were raised there either. I think we can merge this as soon as we get an approval.

[BigLep]

What are the next steps for merging this? Are we waiting to also include #554 or is that going to be a separate PR?

[MarcoPolo]

#554 Will happen in a separate PR. I'd like @marten-seemann to give his approval here before merging