Re: [hybi] BWTP and flow weigth (MulTCP)

[Ingemar Johansson S <[e-mail]>]

Hi

I agree there are two sides to this problem

I believe Bryan Ford did an interesting job in this area, even though his intention was to run the whole thing over over UDP the multiplexing issues are quite the same and it may pay off to look at the problem formulation and also the solution
http://www.brynosaurus.com/pub/net/sst.pdf esp. section 5.6 

Another problem is where websockets compete with non-websockets stuff over the same bottleneck link. One such (possibly evil) scenario is that the user starts an FTP download in the middle of a "websocket" chat. It is possible that this problem should be solved at the layer below websocket as it is I believe a general problem.

/Ingemar

> -----Original Message-----
> From: Jamie Lokier [mailto:[e-mail]] 
> Sent: den 8 mars 2010 23:42
> To: Greg Wilkins
> Cc: Ingemar Johansson S; [e-mail]
> Subject: Re: [hybi] BWTP and flow weigth (MulTCP)
> 
> Greg Wilkins wrote:
> > 
> > Ingemar,
> > 
> > note firstly that it is my intention to come up with a 
> reworked BWTP 
> > as a websocket subprotocol.
> > 
> > But eitherway, I think that having flow control for BWTP channels 
> > would be highly desirable, specially as one of the 
> potential uses of 
> > BWTP is to multiple multiple clients together.
> 
> This is flow weight, not flow control.  They are really quite 
> different, even though they both "control" the flow.
> 
> Flow control is absolutely required when multiplexing 
> independent clients without their cooperation, otherwise one 
> client can block another from making progress, breaking their 
> independence.
> 
> Flow weight is more like bandwidth management or QoS, except 
> I presume 'CC' refers to congestion control.  It does indeed 
> make sense to include it in multiplexing.  Good multiplexing 
> needs fairness heuristics in general too.
> 
> -- Jamie
>