[rmc]

they should use a much better predictor of disadvantage than skin color

What would you suggest? All methods of trying to measure "disadvantage-ness" will be flawed (including race/ethnicity ones). I mentioned this in this comment http://news.ycombinator.com/item?id=4758649

[ta12121]

Yes, they will all be flawed, but not all of them will be as flawed as skin color. Just because something is hard, doesn't mean it's OK to do something easy and bad.

[rmc]

What makes you think skin colour is the worst possible method? Surely flipping a coin is much worse? Or requiring that applications are written in joined-up-writing/cursive?

There are plenty of worse methods.

[ta12121]

I never said it was the worst. I just said there are others that are better. Please.