Re: Reinventing the TransRelational Model?

falcon wrote:
> I have been trying to understand the TransRelational model myself. I
> don't think simply having column based tables make the TR model.
> Sybase IQ has column based table. KDB has column based tables (Dennis
> Shasha calls them Arrables or array tables). MonetDB is not only based
> on 'columns' but there is an actual open source implementation
> (http://monetdb.cwi.nl).

Developed by groups with a very good track record on query optimization and has a lot of peer-reviewed work supporting it. Highly advised, including the associated XQuery work.

> The most interesting research I have seen is something called DODO.
> There are some papers:
> --The Dodo Query Flattening System
> (http://www.ub.utwente.nl/webdocs/ctit/1/0000010d.pdf )
> --A general approach to query flattening
> (http://db.cs.utwente.nl/Publications/PaperStore/db-utwente-42D3A95D.pdf
> )

Agreed, although it's more about showing that a "flat relation" approach might work even if you data is essentially nested. Oversimplifying a bit you might say that the column-based approach is about showing that a "flat binary relation" approach might already be sufficient. :-)

Note, btw. that the groups in Twente and Amsterdam are in close contact. That is not a concidence.

• Jan Hidders