DISTINCT is definitely a bad idea, performance-wise. Avoid if at all possible as it usually involves having to sort the result set and then scan through it, making lots of unnecessary overhead that can often be avoided.