[SteveGarward]

I always encourage students (and trained professionals!) to think through their algorithm on paper/whiteboard, then put the algorithm steps into their code as comments, before writing code. The code then fills in between the comments. If your algorithm was right, your code should match. I find this better than simply commenting the code that was written after the fact - that doesn't make the algorithm right, it just means you understood what the code that was there was doing, enough to document it.

This also tends to lead to documenting the important steps in the code, not every line for the sake of it. Just like code, clarity, brevity, and correctness in comments are probably the most important factors.

(Comments in assembler are a different matter entirely...)