Re: Reinventing the TransRelational Model?

falcon wrote:
> My understanding of the UB Tree (that's the latest from Prof. Bayer as
> far as I know) is that he adds this zig-zag method on top of an
> ordinary B-Tree. The zig-zag method itself seems to be a scheme where
> certain attributes are decomposed into individual bits, those bits are
> then mixed together (first take the first bits from each attribute,
> then append the second bit from each attribute, etc.). There is,
> apparently some mathematical property which shows that these interlaced
> bits strings allow for for effecient searches. In other words, these
> interlaced bit strings allow fairly effecient multi-column indexes.
> Again, this is my very crude understand, I haven't studied this stuff
> in a year or two, so take this (barely coherent description) with a
> boulder of salt :)
>

Yes, it was the Universal Btree that I was thinking of. When I looked at it, it didn't make much sense for my purpose which was memory-based.   Like the ctree, it also depends on a fair amount of cpu horsepower as well as being a 'two-stage search' (my words). However, I was impressed by it, very elegant and it seems that Bayer's imaginative powers haven't been dimished.

p Received on Thu Apr 13 2006 - 02:19:46 CEST