On 7/28/2016 1:07 PM, Olle E. Johansson wrote: >> I.e., we should NEVER use these boxes to govern how we build TCP for the >> > masses. > That’s right - but we do need to find out what happens. Right now our > support is getting a lot of issues that is hard to explain and we’re not even > customers of these carriers. Telling the customer that the net is broken > doesn’t help us - “Facebook, Pokemon and the others work”
It might be useful to have tools that both detect these boxes and "repair" their effects, e.g., by tunneling over them. However, we need to be careful not to build in these temporary repairs into our widely used protocols. > We either break the assumption that TCP is reliable, find these boxes and > refuse to work > with them or will continue to get complains… The boxes have broken the assumption, not you. I agree you need to find a way to undo what the box is, but corrupting TCP to be tolerant of byzantine network behavior is a bad idea IMO. > We are pressed into a corner and want to find a way out. > > At some point we as a community need to find a good “INternet test tool” > for the masses that doesn’t just focus on download bandwidth - but also > quality of the Internet connectivity. Agreed. Joe
