Re: [tsvwg] The List (of application-layer desired features)

Yuchung,

> There is an old work called congestion manager but it's not useful b/c
> it's sender based.

Why does that make it useless? What's missing - isn't it just client=>server signaling of: "I assume the bottleneck is my access link, so I'll tell you which flows belong together, and I'll tell you my priorities"?
But - if you knew which flows share a bottleneck, part #1 would be unnecessary, and then I'm not so sure if you need the client to tell the server about priorities, wouldn't the server know better?

Cheers,
Michael

PS: this makes me wonder about the GreatSnipe congestion manager you once presented in ICCRG - any news about that? Is it going to be a part of QUIC? I think it's very interesting work.

Received on Thursday, 5 September 2013 08:22:55 UTC