Order of layers in MLP

Follow the full discussion on Reddit.
There have been lots of debates about dropout and batch normalization, my question however is, as I do not have much experience with the 2 together nor is my computer strong enough to test without huge time losses, what order should everything be in? To me, it seems most logical if it went: {...}->{(dense layer)->(batch normalization)->(activation)->(dropout)}->{...}; though many seem to disagree.