Robert Scoble has just interviewed the founders of Twitter, given the technical problems which Twitter has been experiencing they seem to be understandably stressed – although playing down the problems as much as possible.

The problems mostly seem to stem from the rapid growth of Twitter plus the traffic generated by having an easy to use plus fairly unrestricted API. It’s understandable for a software house not to worry too much about massive scalability when starting out, that’s time-consuming work for an application that may not take off, however I wish Scoble had asked more about whether they had made any plans to scale up in the early stages of building the application. It would be interesting from a sort of forensic development point of view to know what they had expected to do and what the realities were, once the scaling solution proved inadequate.

The technical problems with Twitter are magnified by the real-time nature of the application, when you expect updates every few minutes (or every second in Scoble’s case) then you really notice service outages, even those which only last an hour or so. Like the guys say, making changes to the codebase in a real-time app is like changing the tire on a car travelling at 95 miles per hour.

No definite timescale has been given for improving the service so for now at least – something is still technically wrong.

Update: June 1 2008

If the details provided in this post by Michael Arrington are correct then Twitter is definitely the victim of poor planning, I know from experience that there is a big difference between application developer and system architect, Twitter seem to be lacking the latter. If the system really does depend on somebody watching the system and manually switching databases when one fails then they simultaneously need and don’t deserve the latest round of funding.