> Do you have a reference, or more information on what a Peano code is?
Check this out http://www.statkart.no/nlhdb/iveher/hhtext.htm
Shortly, this technique relies on a space filling curve. That is, a
uni-dimensional curve that, on a given plane, covers every single point, and
only covers it once. Since the curve is 1-dimensional, one can use B-tree
indexes on it.
There a number of curves of this type, e.g. Hilbert's and Peano's. The Peano
curve yelds easier calculations, hence is the one Oracle used to make their
Spatial Data Option.
Moreover, the Peano curve describes a point in an helical kind of way,
recursively dividing space. That's why the Norwegian Hydrographic Service
decided to call it "Helical Hyperspatial Codes" (hhcodes). It was from their
research that Oracle Spatial Data Option was born, back in 1995.
I'm not sure about the exact applicability of hhcodes to index multimedia
stuff yet (images, sound), because those are VERY high-dimensional spaces.
But I've done quite some reading/research, and hhcodes have two very nice
advantages over R-trees:
- it is easy (and not costly in performance), to index things in 3D or 4D
(including time too);
- concurrency is much better, because one does not suffer from costly R-tree
updates (B-trees are much better in that). When dealing with 3D or 4D, this
becomes even more important.
By the way, are you Brazilian Bernard? Oddly enough, maybe we live in the
very same city. Florianopolis, SC, Brazil. It's a small world he? :))