[OSM-talk] OGL, OSM, NASA
sxpert at esitcom.org
Wed Apr 26 11:52:55 BST 2006
Nick Hill wrote:
> 1) Do geometric R-tree type indexes need 2, 4 or 8 indexed values per
> 2) Do geometric R-tree indexes grow in a non-linear fashion with number
> of indexed rows?
> 3) Does the R-tree indexing time for new points necessarily grow in a
> non-linear fashion with number of indexed rows?
> 4) If the R-tree can't fit in memory, is there an R-tree implementation
> which can substantially cut down the number of disc seeks?
the source to the GIST implementation lives here :
standard data types are here:
> 5) Are there any other relevant questions to using R-trees? If we get
> positive answers to the above questions, we need to re-run tests in the
> light of those answers to check this.
> We than need to compare the results to the possible optimisations we can
> make to b-trees.
r-trees have been specifically designed for 2D things...
> This way, we can save a lot of work for others by mathematically
> establishing the design decisions we must offset for the open streetmap
More information about the talk