Common Challenge

This is way off-topic, but I just couldn’t help myself. Everyone on blogosphere is using Twitter for communication, promotion and (sic!) spamming these days. And, of course, everyone is complaining whenever Twitter is down (which does happen quite often). There were various posts on Twitter’s scalability issues, wrong architecture and even CTO shifts. However, I haven’t read any serious analysis of Twitter’s problems. After all Skype processes a lot more date all the time and it doesn’t break anyhow (I recall just one outage - but a major one…). True - it’s not possible to have a conversation (even one-way) with 5,000 people on Skype, and this is (probably) why Twitter goes down so often. On the other hand, BitTorrent clients can communicate with hundreds of peers simultaneously with no hassle.

And then it hit me: Twitter is archaic. It is constantly trying to uphold the old client-to-server communication, which considering the scale is quite hard. Rather than focusing on scalability and performance of it’s servers it should have a totally different architecture: Peer 2 Peer. Of course - it wouldn’t be web-based then. Of course you’d have API issues. It would be harder to monetize. Maybe it is even impossible to plan. But it’s surprising no one mentioned it before. Maybe I’m wrong somewhere - I’m not an expert in architecture, - so please correct me if you have other opinion. But until no one proves me wrong, let’s face it - Twitter is archaic.

ps. (Unfortunately) After writing this I found an excellent post by Alex Iskold at RWW. He claims that changing Twitter into a protocol would not solve the technical problem (judging by Alex’s experience he may be right on that), but his argument on the looks and ‘feeling’ of twitter are less accurate, as many people use twihrl and other applications to post to and read Twitter anyway.