Re: 'Theoretical' DB OS

On Mar 24, 4:16 pm, paul c <toledobythe..._at_oohay.ac> wrote:

> Like some others here, I'm just guessing at what you mean, but "files
> are attributes" suggests to me that it is something similar to what some
> people call a "column-based store".

Good lead. I'll search and read about this.

ok. So what does this mean? Is it good or bad?

The main point I can see about storing like this is that when you view data it will be fine tuned to the attributes you want to pull in and only those attributes. I can't think of any other good way to do it except at what seems to be the most fundamental part of a database, the attribute.

Are there better ways than using a superkey per attribute item?

The bad part about this is that the size of the superkey is constant and necessarily big. A better way might be to give a superkey to each attribute and each attribute item would have a selectable fixed size under that attribute.

That would lead to stronger typing because you would need to select the maximum number of entries for that attribute to start. This is just like selecting a normal key in a relation and noting it's type/ width.

I was just hoping that there was one single good way to go about this. It probably depends on the problem domain and that's why there are so many different systems and programs out there already.

This is very exciting to me because all the systems I've seen are way more complex and seem more limited than what I expected or need. This also gives me the opportunity to think about WHY there is so much complexity and what appears to be so little return. Fine tuning.

Christian Received on Sat Mar 24 2007 - 23:40:57 CET