If you had to build the current version of Facebook from scratch, which technologies you would use and why?<p>Let's assume you had to build an application of complexity comparable to the current version of Facebook and can take all the time you need.<p>What technologies would you use for each part of the application? Why these? Which languages would you build it in? Any existing libraries or would you write everything from scratch? Why?
I like your question... it's a shame it hasn't garnered more response.<p>Fundamentally, by removing the time consideration the question becomes very unrealistic, but I'll bite anyway.<p>First, you'd have to establish some purpose beyond cloning. Cloning to free the world of a centralized facebook and the surveillance burden that goes with it? Let's run with that.<p>Design-wise, we're looking for adoption. The easiest identifiers to use are email addresses and phone numbers, though neither are particularly secure we can play on the fact that many people have both to offer two factor authentication of public keys while supporting the only-has-one use case with snail mail or alternate communications path (eg. in person key digest validation) recommendations. OK, so now we've got half a basis for some decentralized exchange that supports mass adoption.<p>Next up is sell. How to make this more interesting to the average punter than Facebook, while maintaining global appeal and not having to take on the burden of content creation ourselves? Perhaps with strong support for the content classes people are time and time again proven to care about: drugs, porn and p2p media. This doesn't necessarily have to break any laws, but it probably will in some self-labelling-as-democracy-exporter draconian regime, so we have to include this factor in the design. Luckily, decentralization will be a big help.<p>Finally, we would want to incentivize people to market outside of the network to pull in new friends. This could be done by developing some unique physical use cases like phone-to-phone ad-hoc wifi based sharing and public use local storage drop wifi boxes around major cities.<p>The total hardware weight of such a clone would not be as big as Facebook, simply because of its distributed nature. In fact, it would be far smaller. The search would be different, and would only affect people who chose to register in a central registry. This could even be developed by a third party (or better: multiple third parties, seeded by giving away a proof of concept open source codebase) in order to enhance diversity and thereby strengthen the infrastructure against authoritarian takedowns.<p>Well, just some ideas. Less tech-stack oriented because connecting to fewer peers would probably remove most of the scaling problems. I think that mobile platforms generally demand a certain tech stack, and of course web stacks have been done to death here...
A lot of the tools Facebook uses now are out of necessity because of the massive number of active users. Do I have to build something to support billions of users from day one? Or do I just have to support all of the functionality but just a handful of users? The two scenarios would require vastly different approaches.