I did this once in an interview, but it was a high level "how would you tackle this problem" question. We already had a solution implemented though, so whatever the candidate said wouldn't have mattered much. I think real world problems are far better than the algorithm implementation problems that plague interviewing.