[derefr]

You wouldn't speak those protocols. You'd run your whole web-browser on Earth, and then the Earth<->Mars connection would be a minimum-bandwidth delay-tolerant DOM-event synchronization protocol† to allow you to see what that web-browser is rendering, and tell it to go click things/type things. Essentially like a graphical thin-client protocol (X11 or RDP), but the primitives would be HTML/CSS elements and binary assets in their native formats, rather than native controls and assets in decompressed texture format.

† I believe the Chromecast already operates on some variant of such a protocol. It'd just need to be modified to expect events to come from both sides, rather than only from the "sender" device. (And to remove any/all timeouts and keep-alive pings, of course.)

...or, you could just use https://mosh.org to shell into any random server and run shell commands (including, say, w3m. But even things like IRC would work this way!)