On Aug 19, 2019, at 15:06, Robert Moskowitz <rgm-ietf@xxxxxxxxxxxxxxx> wrote: > > I take Ohtasan’s thoughts with serious consideration. […] > > There are fundemental flaws in our IP stack. We have not tackled them well. I think there are two interesting streams of thought in this thread: (1) We should tweak the mainstream protocols a little bit in an incompatible way so we can get a bit better performance. (2) Maybe we aren’t doing addressing right. I’m seeing the following symptoms here, and I think we should be aware of these. (But I don’t want to topple what’s there, right away, this is really for future things.) I think this thread has thoroughly refuted approach (1), which is essentially an initiation ritual for people learning about interoperability: No, you don’t get to tweak existing, widely deployed things in a fundamentally incompatible way just to scratch a (performance) itch. The angry reactions here are those of people who want to make the Internet great again but run into all these people who try to re-open decisions that already were made. I think we actually are spending some good effort on (2), e.g., LISP. The potentially successful work on (2) does not simply claim there is a higher hilltop with a better view somewhere else, but also tells us how to get from our current hilltop to there. If it can’t be deployed (and that involves creating incentives for all decision-making parties to start the exhausting march to the other hilltop), it may be a great Gedanken experiment, but it’s not very useful as a recipe for evolution. (Overlay networks are a way to stay on the current hilltop and enjoy the other one, too, so the analogy is limited.) Now if someone please created the Final Ultimate Solution to the MTU Problem, next… Grüße, Carsten
