Flakey Twitter and the use of consecutive ids
Twitter was just inaccessible for maybe a couple of hours. Prior to that there was a 9-day gap in their timeline, noticed by at least a few people. I quite regularly have twitters I send not show up at all.
I wonder what could be going on over there? Things certainly don’t feel very stable.
A friend signed up tonight. Using the Twitter API you can see her id. It’s a bit over 10 million. You can also see the id of her first twitter, a bit over 417 million. The earliest twitter available on the system is number 20 “just setting up my twttr” sent at 20:50:14 on Tue Mar 21 2006 by Jack Dorsey who has user id 12 (the lowest user I’ve seen).
Given that Twitter seem to be using consecutive ids for users and twitters, and that you can pull dates out of their API, it would be pretty easy to make graphs showing growth in users and twitters over time. You could probably also infer downtime by looking for periods when no twitters appeared. This would be pretty easy too. Beyond a certain point in time it would be very accurate (i.e., when there are so many twitters arriving that a twittering gap is suspicious), and you could calculate confidence estimates.
I don’t have time for all that though.
But I wonder if Google did something like that as part of their competitive analysis when they decided to buy Jaiku, or if Twitter’s investors did it, and how the numbers would match up with whatever Twitter management might claim. I’ve no idea or opinion at all about any of that btw. But I don’t think I’d be exposing all that information by using consecutive ids for users and their twitters.
You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
November 24th, 2007 at 9:41 pm
[…] have much better and more important things to do than hack on my ideas for measuring Twitter […]
November 24th, 2007 at 9:41 pm
[…] have much better and more important things to do than hack on my ideas for measuring Twitter […]