> We used to just leave that responsibility to the querying host’s own resolver.

When? I didn't get online until the 90s, but resolv.conf delegating recursion to some other node was the way of the world back then. And still is. Some people do run a recursive resolver locally, but it's not so common afaik.

> In any case, if all resolvers these days are recursive why mention it?

Quad9 is named as a recursive resolver in contrast to being an authoritative server. Quad9 and the server I run for my domain are both DNS servers, both provide name resolution, but one is recursive and the other isn't. Used to be, you might run one server for both jobs, but that fell out of fashion for good reasons.

You'd have to go back to the 1980s I think, when systems were still using /etc/hosts and if the host you wanted to talk to wasn't listed there you'd have to use its IP address directly.

IIRC there were some trusted lists of "well known" hosts that were passed around between administrators. DNS was invented because that obviously became impractical quickly as the internet grew.

Back then most computing was done in office, lab, university settings. Running recursive resolvers for the local network was part of the standard IT setup.