The internet is already widely interoperable. But some efforts of late make it less interoperable, and less able to route around problems in the network. The o

The internet is already widely interoperable. But some efforts of late make it less interoperable, and less able to route around problems in the network. The obsession with HTTPS has to end. Not because of privacy, but because of interoperability. DNS over HTTPS is not interoperable with the whole universe of existing DNS products. But what's worse, it's locking people into centralized platforms. With traditional DNS, you can move to any network at all, and automatically pick up a new local, fast, customized DNS catching resolver designed for the network you're on. If you're on DNS over HTTPS, you've always got the same provider, which does not scale to every network. The solution, people will tell you, is just to disable DoH. Until we no longer can, because everything expects to use it. The obsession with HTTPS has also led to the apologists decrying any technical solution that doesn't use TLS 1.3 and HTTPS, because middleboxes!!!! And because literally everyone is reluctant to design new protocols that can be extended as successfully as HTTPS. If it doesn't work over HTTPS, it's not part of the modern internet. This not only severely restricts how you can design technical solutions today, it's stupid: we have this transport protocol with 65,000 port numbers, but we'll only ever use one of them (443), because a redesigned stack is just unfathomable. Every modern network service today needs many things. Routing metadata, dynamic host/service lookup, federated authentication+authorization, encryption, geo-localized load balancing, error correction, session management, etc. If we build things like these into lower levels of the stack, and build primitives for them into the operating system, then all applications can gain their benefits, and we won't need to rely on convoluted hacks to provide it all. We can't keep on for the next 100 years with the shitty protocols and shitty solutions we have today. We have to start thinking about brand new designs, and how we will upgrade systems to use them. Otherwise, every solution we come up with will just become more and more convoluted and ridiculous, as we build more and more on top of antiquated systems designs from 40 years ago. Phone lines were pretty cool. We were able to extend them to transfer data, from 1400 baud to 1.5 megabits. We could technically do up to 50+ megabits, but it wouldn't scale. So we built new solutions. They were expensive, but we needed them in order to grow. Well, I think it's time for tcp/ip and its related protocols to be replaced as well. Not immediately, but it's time for us to start building the replacement. That new replacement can take everything into account in a variety of new stacks. Federation of data, access, services; new kinds of encryption and privacy mechanisms, new trust models. New routing and service models to make the "last mile" less complicated and more flexible. And more responsive to network partition, including the ability to detect them early, to make applications more responsive. We can do literally anything we want, people! We can start building the future today! But we have to choose to do it!