1. 2

  2. 5

    I can’t quite follow the reasoning here, probably because I’ve not worked on this kind of large-scale network engineering. In particular I can’t seem to put my finger on any concrete problem here - just repeated claims that innovation is stifled by needing to reason about how to deal with IP addresses.

    But the solution, near as I can follow it, sounds like having each autonomous system do its own arbitrary, randomized addressing and have another layer built on to map those random addresses to an actual destination server and route within the AS? That seems like adding complexity, not removing it, for all but the largest organizations (which are compelled to build this kind of thing anyway).

    1. 2

      Not sure exactly how Cloudflare does it internally, but each server might have every service installed. That is at least one option if doesn’t want to deal with a lot of internal routing, as you pointed out.