The answer would be simple: Decentralization is a difficult problem.
I suggest you read The fifteen year struggle of decentralizing
privacy-enhancing technology. In this paper the authors list 4 major problems that needs to be addressed for decentralization:
- Incentives in decentralized systems
Most people take up bandwidth more than they donate and proposals to introduce incentives into Tor have failed. . If one would also have to build an incentive system into a decentralized system, there should be a way to manage the ratings of each client without allowing falsification.
- NAT traversal
A truly decentralized system requires the participating nodes to have direct connection to each other. But because IPv4 addresses are getting scarce, ISP often tend to put more subscribers in a single network address suing carrier-grade NAT.
- Bootstrapping new nodes
If in a decentralized system one client wants to become a bridge or relay, there is no central directory server to tell a new node where to locate neighbours.
- Key exchange
With a decentralized model, for a secure communication users will have to share the keys directly with each other making it vulnerable to man-in-the-middle attack.
They also give a comparison of decentralized peer-to-peer overlay networks:
It is interesting to read their conclusion:
..we conclude that there is no fully decentralized system capable of
offering Tor anonymity today. For the first time we document in
detail, the amount of wasted effort and pain spent in
decentralization. The current generation of technology lead by Tor
still has room for improvement, while the next generation is only just
appearing on the horizon. The major problems involving
decentralization are excruciatingly difficult to overcome. None of the
projects have succeeded in making the internet secure and private.
