In the "classic illusions" everyone (or at least the vast majority) of people will describe what they are seeing in a similar manner (an "illusion").
Then, when given an "aid" (such as hiding part of the background) they will "flip" and describe what they are seeing in a different manner ("reality"). There is no such "aid" for the dress (or at least I haven't seen such).
It is thus not an "illusion". It is a segmented perception of reality, where reality is this one specific image discussed.
It demonstrates that given a specific set of colors on which there is no argument (see wired RGB image), a composition of those same colors can still be perceived differently, but consistently, by segments of the population. They see a different "truth".
BTW my 10 people sample had 5 blue eyed people in the black/blue camp and 5 green eyed people in the white/gold camp.
http://web.mit.edu/persci/people/adelson/checkershadow_illus...
On another note, I'm wondering if people did comparisons on whether the size at which you viewed the robe had any effect on people's choice. Personally I can't see it as white because there are enough cues in the background that indicate that there isn't a strong blue light lighting the dress from the front. On the other hand if I shrink the image and squint, I can imagine it being a sort of backlit photo where the blue tint would just be a result of outdoor atmospheric diffusion and the contrast with yellow sunlight.
Take this example... http://assets-s3.usmagazine.com/uploads/assets/articles/8371...
Now does everyone seeing that see bronze and light blue on the left, and dark blue and black on the right?
Supposedly these are images of the same dress. If anyone suggests they're seeing the colours differently on this example, I'll be genuinely surprised, but at the moment I find that unlikely.
EDIT: ...and the downvotes have started. How predictable.