you may not now how geo index work for 2d space well if u don't look here http://docs.mongodb.org/manual/core/geospatial-indexes/ With those index you can efficiently get all the data in a square.
Now we gonna try to get all the data that are in a n-dimensionnal cube and use the same principal as the geo index but in a space of dimension n. Why ? Well first case you have a large pool of offer maybe mobile abonnement, services or even a dating site and you want to give them the closest offer to what they want, not the exact offer because it may not exist but the closest. For example i want a mobile abo with iphone 5 or better phone, for 40euro a month and 12 month period. We can put each criteria on an axis even use some weight on the data and then just use a geo n dimension index to get the closest offer.
Continious axis no weight our test impl would be the time to go to a shop and the price of the offer and time
Just realized this was called a R-tree
handle discrete axis with a limitied amount of step