Nothing, but I think the implication is that it is a signal that can be used to predict which of the three categories the person falls into. People who fall into the third category (the desirable one) are likely to be able to implement mergesort on a whiteboard because they have a genuine interest and paid attention.

That being said, other, more practical questions may exist that retain the signaling characteristic.

With respect, I'd like to add a fourth category:

4) Actual computer scientist. If you ask him how to write mergesort, he will ask for a book to look it up in, same as if Einstein was asked for the value of pi. If you ask him to describe his favorite algorithm or the details of his current research, he will gladly tell you in extensive detail.

The problem with hiring those of us in category (4) is that it's hard to tell if we're bullshitting you, since after all we operate at a far more advanced and specialized level of expertise than most software development jobs actually use or require.

Sorting algorithms in general aren't constants, though. The implementation can have significant impact on running time, memory use, etc. And there are an infinite number of them with various tradeoffs you can tune.

I wouldn't expect an interviewee to know sorting algorithm X because they may just have forgotten. But why wouldn't they be able to come up with a sorting algorithm, and have some feel for the behavior? Seriously, selection sort is about as easy to come up with as fizzbuzz. If they can't think through a simple problem and come up with a solution, is it wrong to question their problem-solving skills?

Actually, if you asked in an interview how to sort, I'd try to come up with quicksort or mergesort. I wouldn't give you insertion sort or selection sort because I'd feel terminally embarrassed to go with so bad an answer.

Is it terminally embarrassing to say "this is bad, but..." and pass the test, then discuss (maybe write out) more advanced options? If all they care about is selection sort, you've saved time (often a good quality). If they care about more, you've demonstrated something, and opened the door for more.

Depends. I've been in interviews where they let me damn well know I was losing points for giving an algorithmically sub-optimal answer.

Sounds like (a) fun interview(s) :/ Alas, they exist. I guess it's too much to hope they don't.

Where did he say he was writing web software?

The author of the parent post probably assumed it from radicalbyte's "about" section.