An excerpt from the best rant about whiteboard interviews posted on the internet. Ever.

"Well, maybe your maximum subsequence problem is a truly shitty interview problem. You are putting your interview candidate in a situation where their employment hinges on a trivia question. — Kadane's algorithm! They know it, or they don't. If they do, then congratulations, you just met an engineer that recently studied Kadane's algorithm.

Which any other reasonably competent programmer could do by reading Wikipedia.

And if they don't, well, that just proves how smart the interviewer is. At which point the interviewer will be sure to tell you how many people couldn't answer his trivially simple interview question.

Find a spanning tree across a graph where the edges have minimal weight. Maybe one programmer in ten thousand — and I’m being generous — has ever implemented this algorithm in production code. There are only a few highly specific vertical fields in the industry that have a use for it. Despite the fact that next to no one uses it, the question must be asked during job interviews, and you must write production-quality code without looking it up, because surely you know Kruskal’s algorithm; it’s trivial.

Question: why are manhole covers round? Answer: they’re not just round, if you live in London; they're triangular and rectangular and a bunch of other shapes. Why is your interview question broken? Why did you just crib an interview question without researching whether its internal assumption was correct? Do you think that “round manhole covers are easier to roll" is a good answer? Have you ever tried to roll an iron coin that weighs up to 300 pounds? Did you survive? Do you think that “manhole covers are circular so that they don’t fall into manholes” is a good answer? Do you know what a curve of constant width is? Do you know what a Reuleaux triangle is? Have you ever even been to London?

If the purpose of interviewing was to play stump the candidate, I’d just ask you questions from my area of specialization. “What are the windowing conditions which, during the lapping operation on a modified discrete cosine transform, guarantee that the resynthesis achieves perfect reconstruction?” The answer of course is the Princen-Bradley condition! Everyone knows that’s when your windowing function satisfies the conditions h(k)2+h(k+N)2=1 (the lapping regions of the window, squared, should sum to one) and h(k)=h(2N−1−k) (the window should be symmetric). That’s fundamental computer science. So obvious, even a child should know the answer to that one. It’s trivial. You embarrass your entire extended family with your galactic stupidity, which is so vast that its value can only be stored in a double, because a float has insufficient range:"

Author: John Byrd
Src: https://quora.com/What-is-the-harde...