[mathogre]

JFC, that's such bullshit. You want the mundane? Go for it. You actually want someone who can take a bunch of real operational data and solve the problem when the Person With The Money says, "We need to know what is actually happening with ____. And we've promised it in two days. You're it."

Who gives a flying fuck about writing the best sorting algorithm? "sort" works just fine, unless you're Google and microseconds matter. And then, you're really at the edge of R&D. You need to be able to manipulate data with aplomb. You need to be able to write an algorithm that works, and then refine it to make it go a hundred (or more) times faster once you understand why it is so slow.

[hawkice]

I'm not a Googler, but I imagine memory usage would be a more meaningful constraint than speed per se on the mega-sorts they are doing. Completely speculative, but in service of a point valid even if the statement isn't: in those rare circumstances, the needs of the company are going to bizarre _anyway_. If you don't know what demands you'll face in the future, there's limited value in learning ahead of your needs.