Minutes for HyBi session at IETF 84, Vancouver
Thanks to Patrick Mcmanus for additional notes.

THURSDAY, August 2, 20121300-1500  Afternoon Session I 

PerFrame Deflate draft - Takeshi Yoshino
========================================
http://tools.ietf.org/html/draft-ietf-hybi-websocket-perframe-compression/
    
Issues with multiplexor quotas and per-frame compression (needs to uncompress, re-fragment and re-compress to fit quota).
		
Easier to compress per-message and not per-frame.
		
Per-frame interleaves messages in the presence of mux.
Per-msg compression is not applicable after multiplexing.
		
Proposal to do things on a per-message instead of a per-frame basis:
* Per-msg compression
* Mux by encapsulation - instead of extension data on a per-frame basis

Changed compression ID name to "permessage" from "perframe".
		
Positive sentiment in the room.

[sal] can we adopt this document with the name change?
[barry] that's up to the chairs and wg
[gabriel] took sense of room and got 2 thumbs up and no dissent - off to mailing list to confirm]
        
Multiplex draft - Takeshi Yoshino
=================================
http://tools.ietf.org/html/draft-ietf-hybi-websocket-multiplexing/ 

Encapsulation solution for mux
* Into a binary msg with channel ID (one more byte)
* Easier to do post-mux compression

Updates on flow control for mux.

Decoupled compression algorithm and framing. Initially assuming deflate, but allow for different algo's in the future. Also, compression algorithm is echoed back to simplify the negotiation.

TODO:
* Defend against attacks by having a send quota on control frames
* Define mux error codes
* How to fall back to another physical connection

Takeshi also talked about mux and HTTP/2.0, and of possibility of using WS 1.0 as a basis. Some warning from the room that it is important to continue with WS 1.0 including mux and compression, so this work should continue.
            
[gabriel] websocket basis for http/2 likely not to happen in his opinion: this was one of the proposals for HTTP/2.0, but HTTPbis decided the previous day for a different basis. Current idea is what Brian will present next.
[fette] websockets over spdy/http-2 will come later, so keep it in mind.. but we need something for WS 1.0
[mcmanus] flow control penalty should be proportional to bwidth not 1. 
group can wind down with completion of this extension definition


HTTP 2.0 and WebSockets Coordination - Brian Raymor
===================================================

Idea is that HTTPbis will work on HTTP/2.0, including mux and framing for 2.0. They should work on that and not impose artificial limits, so protocol reuse will be possible with a future version of websockets ("websockets 2.0").

This has been added to the tentative HTTPbis charter's text on HTTP/2.0 work.

[fette] is http/2 good for carrying both ws and http on same connection 
[brian] I think that would be a requirement 
[gabriel] new httbis charter has been modified to add directive on coordination with hybi 
[fette] mixing http/2 is a must 
[mcmanus] really a httpbis item 
[gabriel] just asking them to coordinate with hybi [mcmanus] hybi should cease to exist when charter is complete 
[gabriel] have httpbis do work to ease the road for hybi future work