Decentralized systems

All decentralized systems must be distributed. But distributed systems are not necessarily decentralized. This is confusing to many people. If a distributed system is one spread across many computers, locations, and so on, how could it be centralized?

The difference has to do with location and redundancy versus control. Centralization in this context has to do with control. A good example to showcase the difference between distributed and decentralized systems is Facebook. Facebook is a highly distributed application. It has servers worldwide, running thousands of variations on its software for testing. Any of its data centers could experience failure and most of the site functionality would continue. Its systems are distributed with fault tolerance, extensive coordination, redundancy, and so on.

Yet, those services are still centralized because, with no input from other stakeholders, Facebook can change the rules. Millions of small businesses use and depend on Facebook for advertising. Groups that have migrated to Facebook could suddenly find their old messages, work, and ability to connect revoked—with no recourse. Facebook has become a platform others depend on but with no reciprocal agreement of dependability. This is a terrible situation for all those groups, businesses, and organizations that depend on the Facebook platform in part or on the whole.

The last decade has brought to the forefront a large number of highly distributed yet highly centralized platform companies —Facebook, Alphabet, AirBnB, Uber, and others—that provide a marketplace between peers but are also almost completely unbeholden to their users. Because of this situation, there is a growing desire for decentralized applications and services. In a decentralized system, there is no central overwhelming stakeholder with the ability to make and enforce rules without the permission of other network users.