I watched this talk last night. It was well done and I like the overview style Carruth took with the presentation.<p>This is the style of programming I do for basically everything so I was very sympathetic to moves towards a data oriented compiler design. However, there is going to be some grumbles that this does not fit into the 'standard' compiler implementation theory and that abandoning such a previously well researched is foolhardy. I fundamentally disagree, but I'm sure any forum arguments will be fun to watch.