[weberc2]

I would be wary of basing those choices on these data. If I just learned about the concept of programming languages for the first time, I might refer to this for a vague understanding of which languages are popular--I might not opt to learn Coq over JavaScript, but it's not obvious from these data that picking JavaScript will be of greater benefit to my career, interests, or organization than Scala. This seems like one metric to consider, and not even a very good one. In particular, popularity on StackOverflow should probably count against a language, but these rankings consider it a virtue.

[kodfodrasz]

SO popularity is neither clearly positive, neither clearly negative.

Why not positive: because if there are lots of questions about something, it must be very hard, with lots of gotchas, unclear points.

Why not negative: because if there are lots of questions about something, it must be very popular, so many newcomers have their trivial questions, which do not necessarily indicate problems about the language.

[weberc2]

Right, the SO value is approximately the product of popularity and difficulty; the GH value is approximately popularity, so if you difficulty is approximately SO/GH.

[kodfodrasz]

If this would be true, then "XML" and ASP would be the hardest languages. Also this would suggest that TeX for example is simpler to use than Python or Ruby.

I do not think this stands.

[weberc2]

Hence, "approximation". In particular, XML isn't a programming language, so questions about it are more likely to pertain to writing parsers and the like. TeX users may well have other sources for answering their questions. The approximation is very rough, but there is clearly a signal in the noise.