If I know anything about the future, it doesn't look like the present.
The web won't look like it does now in 2050, and neither will the internet.
But it might very well be built on webassembly on browsing engines cum operating systems on top of hypervisors on top of verified microkernels, and the web will probably be delivered on top of HTTP/2 on top of TCP/UDP and so on. The layers probably won't change that much.
If someone solves the multicast problem during that time span things might change radically on several layers. And even if not some immutable, content-addressed protocols might make some inroads. SRI is planting the first seeds in that direction.
I'd venture to say that few people are using the emacs of 1980, and that telnet is mostly confined to legacy applications (not that it means the point is wholly invalid).
The web won't look like it does now in 2050, and neither will the internet.
But it might very well be built on webassembly on browsing engines cum operating systems on top of hypervisors on top of verified microkernels, and the web will probably be delivered on top of HTTP/2 on top of TCP/UDP and so on. The layers probably won't change that much.