DNS is a singular centralized namespace.
HTTP(s) bakes an authoritative server into the name of every object.
Sure there are counterexamples - TCP doesn't require router state, UDP lets users skip the OS protocol stack, DNS could be even more centralized.
But the basic protocols weren't really designed to resist much of anything. Their decentralization arose out of administrative concerns (site A should be able to administer its own names without coordinating with B and C) and engineering concerns (E2E principle) rather than power relations (powerful C can force its will onto the relationship between A and B). Designing to resist against the latter takes public key cryptography, which was nascent and too computationally expensive prior to the past few decades.
IP addressing is not centralized.
DNS has a distributed common namespace with lots of delegation.
HTTP is peer to peer.
The easiest book to read about the personalities and history of this stuff is probably "the innovators" by walter isaacson.
If I compare the networking we came up with to the other types of networks available in the day the ip protocols were invented, I think what we have is remarkable.
Surviving nuclear war isn't quite the same thing.
Although there have been some less-than-optimal implementations in some protocols, in general the internet protocols were designed by people with their eyes open.
The protocols were designed to resist centralized control, and have been very successful.
You should read specifically about how tcp/ip was developed, not technically, but the people behind it and their motivations.