Re: satisfies algorithm

[Snipped]
> In nearly all of the commercial implementations and implementations of
> systems built with commercial implementations that pervade the industry,
> where there is a unique constraint or a primary key constraint that needs to
> be enforced, there is also an index. But thank you for pointing out that
> there are alternatives to using indexes without also providing the
> algorithms for those alternatives for the OP's edification. In other words,
> "where's the beef?" Can you back up your claims with more than marketing
> hype?
There is really *no beef* but simply the application of sound relational concepts onto determining a computing model and encoding scheme that preserves physical logical independence. The fact that there is no need for indexes is just one of many good side effects of the underlying research effort that started many years ago.

I would like to point out that I brought up the example of the db core I am working simply in good faith to support David's underlying claim that physical data independence (and therefore index independence) is anything but utopia. I am not planning making money out of it since it will be open source once the core is completed and relationally sound. I plan on documenting and making available to public the computing model, encoding scheme and algorhythmics only at that point. I am still working out the issues of temporal data and missing information without use of NULLS.

> I read the white papers Netezza provided on their web site, but I
> didn't feel a thrill going up my leg, if you know what I mean. Can you
> provide links to academic papers that describe these alteratives in detail.
> I'm curious, and I would bet that others here are too.
Products such as Netezza are simply an old idea rediscovered. Google up *column store*. For more information about Netezza, contact Netezza people.

> By the way, how Netezza is implemented internally is anyone's guess. They
> may sort active domains--storing them in a particular order. Who knows?
> but wouldn't that be similar enough to be considered a facsimilie of an
> index?
Depends how you define an index. `Direct image indexing schemes are primarily based on the concepts of address pointers associated with some kind of querying scheme(bitmap, b-tree) to allow the system to determine as fast as possible (as little logical operations) where the data is on disk.

> --Brian
Received on Sun Jul 27 2008 - 12:35:11 CEST