Every large web site uses a load balancer, effectively a proxy. I don't understand why Poul Henning-Kamp's experience with that didn't carry more weight in the working group.
In HTTP terms a reverse proxy is just a Web server (what it does internally is its own business), not a proxy. HTTP/2 load balancers have no trouble terminating TLS.
If you disagree with PHK, you have to refute the details in his notes, not just wave them away. That they can be made to work does not mean there is not a problem.
PHK keeps calling for radical changes to HTTP semantics which other people don't have the stomach for and would take a decade to deploy even if they did. In some sense, getting HTTP/2 done quickly clears the field for people like PHK to start on HTTP/3 sooner.
Comments
That's simple, they just won't, browsers will only support HTTP/2 over TLS, so proxies are effectively dead.
Every large web site uses a load balancer, effectively a proxy. I don't understand why Poul Henning-Kamp's experience with that didn't carry more weight in the working group.
In HTTP terms a reverse proxy is just a Web server (what it does internally is its own business), not a proxy. HTTP/2 load balancers have no trouble terminating TLS.
If you disagree with PHK, you have to refute the details in his notes, not just wave them away. That they can be made to work does not mean there is not a problem.
PHK keeps calling for radical changes to HTTP semantics which other people don't have the stomach for and would take a decade to deploy even if they did. In some sense, getting HTTP/2 done quickly clears the field for people like PHK to start on HTTP/3 sooner.
Not entirely true. PHK outlines some specific issues with HTTP/2 that makes a load balancer's job harder and/or less performant.