I think there are a number of better ways to determine which programming languages are popular than what Tiobe uses. If they were only counting pages generated within a specific timeframe, perhaps a month or year, then their results might have some validity. Every month they are including all of the web pages that exist. I believe this skews their results.

I'm also not convinced that language popularity is an important metric. I think knowing which languages are strongly sought after by employers is more important and relevant. I think it is also important to know which languages are suitable for particular tasks, for people looking to create new code bases.

It would also be interesting to know how much code exists in various languages, and how many people are currently employed maintaining code bases in various languages.