The WWW, as originally designed, centers on URLs, which are designed to describe how to access data at “well-known” locations, owned by fixed entities.
This arrangement causes durability issues, because the entities maintaining the servers at those addresses shift and decay. The arrangement also introduces privacy issues, since whatever organization controls that server also necessarily knows about all your traffic to it.
These are the classic distributed web arguments—a mixture of libertarian and archivist tendencies. I’m interested in a weird different frame, though: escaping maintenance responsibilities! If I want to experiment with a web-based application which other people can use, I’m implicitly committing to running and maintaining servers and that domain name, indefinitely. Even if I open source the system, it doesn’t help others very much, since they’d need to deploy and maintain their own parallel servers if they want to run their own! And then there’s potentially a challenging data-porting story between the services. I think this state of affairs suppresses experimentation and iteration.