Browser developers (and by extension, web standards developers) appear to be extremely reluctant to work at any protocol level other than HTTP. To this aversio

Browser developers (and by extension, web standards developers) appear to be extremely reluctant to work at any protocol level other than HTTP. To this aversion I attribute several pathological outcomes that affect many of us. Layering-violation constructs like WebSockets, failure to use SRV leading to hacked-up apex DNS A record tracking, timeouts/misbehaviours giving IPv6 a bad name, the general reinforcement of HTTP as an inefficient and poorly specified RPC transport. And, yes, ignoring DANE. Fun fact: even using DNS A records for endpoint name resolution in HTTP is a convention; it is not mandated in the document body or in any of the normative references. My greatest disappointment in HTTP/2 is that it reinvents multiplexing. Not so much that it wasn't necessary (it was, at least it was if you are Google or Facebook or Twitter scale), but that it occurs inside the protocol rather than in a separate layer that other future protocols could use. To me, that's exemplary of how HTTP has been developed despite, rather than in concert with, other protocols.