Re: Date and McGoveran comments on view updating 'problem'

On Dec 12, 3:34 am, JOG <j..._at_cs.nott.ac.uk> wrote:
> Ah, of course, the header. Now it makes sense - distributivity
> requires correspondence in header unions. It might be worth noting
> that I did not make the assumption that headers intersect when the two
> components of a join are disjunct (and on a quick reglance of your
> work could see no mention of this). In "First Steps in Relational
> Lattice" you very neatly define a natural join over two binary
> relations as:
>
> A(x,y) ^ B(y,z) = {(x,y,z) | (x,y)EA & (y,z)EB }
>
> However, this makes no mention of headers at all, leaving ambiguity
> that might be worth just rectifying in future papers (to avoid any
> confusion in those like me who don't assume a theoretical requirement
> for header constructs).

I never realized that this set definition might be a stumbling block! Jan was criticizing this section of the manuscript as well. My line of thought was: "OK, everybody knows what natural join is, we don't need set definition for that. Inner union, however, is a new construct, so it has to be defined (in set notation). In case if a reader has a problem with inner union definition why not to provide a natural join definition? Then the reader can consolidate definition for familiar operation with his notation confusion". It looks like this idea didn't work:-)

> Nevertheless all very interesting stuff Vadim! Have you formally
> published your work in this area?

Alas, just missed PODS deadline
http://www.sigmod09.org/calls_papers_important_dates.shtml by couple of days! Received on Fri Dec 12 2008 - 19:31:09 CET