On 3/7/2014 3:31 PM, Matt Mathis wrote:
My point is that if losses are small enough where the average goodput is is controlled by the application (and not CC) then the retransmits needed to repair the losses INCREASE the load.
No disagreement there, but IMO that just means that your reliability algorithm is interfering with your ability to enforce congestion control - and I'm not surprised it happens in TCP.
But I don't think we need to use new terms for it, and I don't think "is there CC?" is the wrong question. It's the right question, but often has the wrong answer.
Joe
