Wednesday, March 26, 2014

Twitter's Single Point of Failure

Dave Winer has repeatedly pointed out that Twitter is, in itself, a single point of failure since it's centralized. No one can argue with that. But, the upside is that it's fast – very fast. The only noticeable latency is the time it takes for a tweet to be uploaded from a client to Twitter's servers. When considering distributed infrastructures like SMTP and DNS it's obvious how much slower they are due to propagation. The upside of these fundamental Internet services is no one owns them which means no one can inherently stop them or advertise against them, yet they're cheap enough that companies can provide them for free. But, just because no one owns these protocols, doesn't mean they can't go down, at the end points.

So, it begs a couple questions. Is a fast, free, Twitter, better than the alternative? More importantly, is Twitter a key, fundamental, Internet service worthy of an RFC?

Author: Joe Moreno

No comments: