in [part I](http://www.ebaytechblog.com/2015/05/01/fast-approximate-logarithms-part-i-the-basics/ http://www.ebaytechblog.com/2015/05/01/fast-approximate-loga..

in [part I](http://www.ebaytechblog.com/2015/05/01/fast-approximate-logarithms-part-i-the-basics/ http://www.ebaytechblog.com/2015/05/01/fast-approximate-loga...) the author argues against using [1,2) because it causes large relative error near x=1 (since log x = 0 there). I would disagree (why would you ever care about relative error in a logarithm -- IMHO absolute error makes much more sense... unless for some crazy reason you're dividing by log x near x=1) but that's his argument.