Open ghost opened 6 years ago
@Stebalien I'm closing this in favor of #6. But if you have a second topic for protocol design discussion that you want to lead, just reopen.
(I wanted a full protocol design and proposal before filling this in but, in retrospect, that would have been overkill and completely unnecessary).
Awesome - this is a great topic. Glad you thought of it.
Please drop your notes to https://github.com/libp2p/developer-meetings/tree/master/notes
[Protocol Design] Protocol Negotiation 2.0
Name: @Stebalien
Length (choose one): 30m: 5 minute presentation, 25 minute discussion
Title (~1-7 words): Protocol Negotiation 2.0
Abstract (1 sentence up to a couple paragraphs, as you prefer):
Multistream select is currently inefficient both in terms of round trips and bandwidth, handles things like TCP's simultaneous connect poorly, and isn't usable for packet connections. Incidentally, the bandwidth inefficiency discourages long protocol names like
/ipfs/QmId/path/to/my/protocol/RFC
. Supporting names like this is a goal of the protocol.I plan on presenting a new protocol negotiation protocol and hope to present an upgrade path (but that'll probably take some discussion).