Wired recommends the Baby Name Wizard for tracking the faddishness of names. Names showing a sharp spike upwards in popularity disappear similarly quickly; those instead building slowly maintain popularity over a longer period.
I can also recommend the Wizard and the back end from which it presumably draws, the Social Security Administration's baby name wizard. We, or rather I, made extensive use of both about fifteen months ago. I compiled the set of names that made it into the top 1000 from 1880-1930, deleted any name showing up in the top 100 in any year from 1970 onwards, deleted any names now with reversed gender connotations (Marion, for example), deleted any that were just too anachronistic, added in a few that seemed fun, then picked my top 100. We iterated down to a top twenty by sequential vetoes (Sue vetoed 10, I vetoed another 10, and so on). After our little guy arrived, we ran a Borda count on the 20 to narrow down to a top five, then just matched up by consensus the name that seemed the best fit for the little guy featured in the picture top right. I was mildly surprised to find out that most folks don't use this method for choosing names: the rest of you seem to be choosing sub-optimal algorithms for baby name selection. Now that you know of the optimal algorithm, ignorance can't serve as excuse. It certainly helps in avoiding choosing faddish names.