Beneath IP are things that are hardware, or very close to it.
Hardware physically pushes bits on a medium and gets them off. Hardware is the only layer that can even try to provide any guarantees about those bits, since it's the one physically handling them. Anything above isn't "real" until that point.
But it can't--after all, power may go out, your cat may chew on the Ethernet cable, your ISP cuts you off because your bill is unpaid, etc.
Kinda like the same way your filesystem really wants to treat the underlying storage devices as reliable, but does so at its peril because when a drive wants to die, it's going to die and you're not going to stop it. So you put layers between the high-level filesystem and low-level devices to abstract it.
IP is an internetworking protocol that enables routing. Routing is supposed to be where "RAID" for Internet is supposed to live - ideally there is one global Internet, not millions of balkanized mini-Internets behind NATs, and robust networks would have working multiple routes out of their network, and those routes would be known to all hosts on that network and easily be communicable to adjacent networks. So your path to a given IP could take one of many, and that is what happens between service provider networks.
DNS does suck, we could go back to passing hosts files around. Essentially that's what's happening in reverse with ad-blocklists.
Hardware physically pushes bits on a medium and gets them off. Hardware is the only layer that can even try to provide any guarantees about those bits, since it's the one physically handling them. Anything above isn't "real" until that point.
But it can't--after all, power may go out, your cat may chew on the Ethernet cable, your ISP cuts you off because your bill is unpaid, etc.
Kinda like the same way your filesystem really wants to treat the underlying storage devices as reliable, but does so at its peril because when a drive wants to die, it's going to die and you're not going to stop it. So you put layers between the high-level filesystem and low-level devices to abstract it.
IP is an internetworking protocol that enables routing. Routing is supposed to be where "RAID" for Internet is supposed to live - ideally there is one global Internet, not millions of balkanized mini-Internets behind NATs, and robust networks would have working multiple routes out of their network, and those routes would be known to all hosts on that network and easily be communicable to adjacent networks. So your path to a given IP could take one of many, and that is what happens between service provider networks.
DNS does suck, we could go back to passing hosts files around. Essentially that's what's happening in reverse with ad-blocklists.