We've cooked up a user-mediated form of redundancy in our federation. This makes storage more available but less reliable. I riff on this after Jon Richter observes similarities between our inventions and Amazon's.
Today it came to me the Transporters are what the industry calls serverless computing, where so-called lambda functions (?) are called remotely over the network from within an application's local call stack (if existing) and provide deterministic responses, what they may call lambda computing. You push a function to a, let's say, transporter chassis, and it is available from within a wiki's transporter instantiation screen. A roster, directory of transporters comes to mind for discoverability of available ones.
This is an interesting observation. Amazon's Lambda and our Transporters have no need to remember anything. That gives them nice properties in a networked world. Transporters are especially interesting because their transformation is from something on your screen to something else on your screen. You can bookmark the source or you can transport and fork the result. You are in charge of your own memory.
A design feature of the early web was that servers were "stateless". Database backed sites became popular which broke this model. (Wiki was early among them.) But the distinction stands in the large sites. To get internet scale the web server and the app server behind it have to be stateless so that they can be redundant and load balanced. This leaves someone "holding the bits". Big sites push this back further into complex storage engines. Wikipedia is no exception. But federated wiki does take another path hoping that end users can save and backup their own work and the best stuff will survive the occasional loss through social replication.
Strangely, as the cost of storage drops, the cost of remembering climbs. An author that wants to write for the future has few choices. The cost of administering data clusters greatly exceeds the cost of the media and is hard to fund with advertising except for the most clickable content. If we think of patterns and pattern languages as just more memes struggling to survive then the outlook is bleak.
Yes, this is why we parallely try to establish a computational commons which can hold the most precious and most vulnerable information from everyday livelihoods that don't have corporate or state backing. The latter are also just self-stabilising mnemotechniques, which enforce structure over flow. Why commonify's work is so important.