Re: The TransRelational Model: Performance Concerns

"Josh Hewitt" <lajos.nagy_at_gmail.com> wrote in message news:1c92edeb.0411102354.45157060_at_posting.google.com...
> Recently there has been some excitement created amongst database
> enthusiasts (though not researchers) by the patented technology called
> the TransRelational(TM) Model. It was brought to the attention of the
> general public by C.J. Date who gave a brief description of it in the
> appendix of the 8th edition of his classic book "An Introduction to
> Database Systems". Elsewhere he talks about it as "a radically new and
> dramatically different technology" which would let us build an
> implementation of the relational model which "would be orders of
> magnitude faster than, and would deliver a far greater degree of data
> independence than, today's SQL products."

One of the things turning people off to "TRM" is the patent. Patents discourage adoption or analysis of an approach, because a patent is a substantial barrier for use. Commercial and open source projects both often steer well away from patented technologies, unless an agreement with the patent holder is secured in advance.

As I read over your post, I couldn't help but notice that it looked just like a description of a fully inverted index. What do you think? I haven't read the patent, but I note that the inverted index is a fairly old idea. A quick glance on citeseer showed references to the term from as early as 1988, but I believe it is much older than that; I think it may even date from the 1960s.

Your analysis looked pretty good to me. Particularly telling is the cost of the insert. But one thing you didn't take into account was the cost of a join. I note that if all columns are sorted, then every join can be a merge join, which is darned fast. O(m+n). Dunno if this would make up for the other inefficiencies you noted; maybe not.

Marshall Received on Fri Nov 12 2004 - 01:06:02 CET