This will evolve into a longer one but I wanted to get the essentials captured before they floated away.
There’s a lot of talk today around “decentralized” software and leveraging things like public API’s and blockchain technology. This is cool stuff, but it’s surprising to me how many people excited about this fail to realize that the problems we have today with centralized services on the Internet is a rather new occurrence.
Basically before The Web, everything on the Internet was decentralized. The reason this was possible is because of standard protocols. This might sound obvious, but if you look at how we value an open or public API, it’s value pales in comparison to an open protocol. A public API allows you to interact with a service as it was designed by someone else, but a protocol allows you to interact with any service in the same way. It liberates both the service consumer and the creator from having to worry about compatibility. It also delegates some of the hard questions about how services should interact to larger groups of experienced engineers concerned with the global scope of the Internet vs. smaller groups concerned with (or at least only most familiar with) their own domain.
Let me give you an example. There are an assortment of services that provide file storage in the cloud: Amazon S3, Microsoft Azure, Dropbox, etc. Most of these provide an open REST-based API. However to interact with them each one does things slightly differently, even though they all work over HTTP and attempt to adhere to REST conventions. The result is that if you write code against one and decide to use another (or are forced to switch) you have to write more code that looks a lot like the code you wrote in the first place, but not exactly. Moreover, as each vendor improves their service, new versions of the API are released. If you want to support more than one service, you have a lot of work cut out for you.
Now consider email. Before most people had access to the Internet a lot of them used email, but they were private, closed systems that could only exchange mail within their own network. Some mail systems engineered “gateways” that could talk to foreign systems but these were complex and certainly not universal. The development of the SMTP protocol allowed these systems to talk to each-other without having to continually adopt the other’s changing standards. Not only this, but it allowed mail servers and clients to no longer be tightly bound to each-other, giving users a lot more choice and forcing down the cost of email.
Imagine if the same was true for storage. Imagine if there was a SSTP (Simple Storage Transport Protocol) that you could code against and know your code would work against any compliant server. This doesn’t magically force vendors to adopt the standard, but in the case of email, not having SMTP compatibility put your server at a disadvantage against systems that did, and users demanded it as well (they wanted to be able to email their friends on different systems).
Remember I said this was going to be short? Whoops.
To wrap it up, if we really care about returning the Internet to a distributed system we need to turn our attention to developing new open protocols instead of focusing on implementation, because without these protocols, we are destined to simply create new and more interesting cages.
--
// jjg