[Galene] Re: using up more ports in ipv6 for better congestion control

[Sean DuBois <sean@siobud.com>]

On Fri, Jul 16, 2021 at 03:36:31AM +0200, Juliusz Chroboczek wrote:
> >> Tim, I'd be very grateful if you could explain what advantages TWCC has
> >> over REMB.  For now, I'm sticking with REMB.
> 
> > With TWCC the sender knows the metadata of lost packets. If you lose a packet
> > with REMB you don't know the send time or the size of the packet. That
> > seems like it could be useful information?
> 
> I understand that TWCC is more chatty, and sends more detailed information
> to the sender.  What I don't understand is why this information is useful:
> REMB performs the exact same computation as TWCC, but it does it on the
> receiver side, and only sends the result to the sender, thus avoiding the
> chattiness but yielding the exact same result.
> 
> What am I missing?  What exactly does TWCC buy you?
> 
> -- Juliusz

Receiver Side BWE can't know the size+send time of lost packets.

I am not aware of any other reasons though. In the GCC [0] paper it
looked like calculations were designed to happen on both ends. Maybe it
was more maintainable to have all the logic in one peer?

[0] https://www.aitrans.online/static/paper/Gcc-analysis.pdf