Re: Flamewar object databases vs. relational databases

> Your responses seem to lack any balanced view, so it's meaningless for me to
> discuss this issue with you.

If you say so.

> Suffice it to say, I was pointing out to
> another poster that relational databases aren't the answer to all data
> storage problems any more than a saw is the only tool needed for carpentry.

Well, data storage and database management are different things. I use both hierarchic (DOS) and network model (Unix) file systems. Those are file systems and make no claim to be database management systems.

For storing configuration parameters, I use both flat ascii files (.ini) and a hierarchic database (Windows Registry). I have no end of problems with the registry. I could repair a .ini file using "C:>copy con app.ini" when I absolutely had to.

I store partition information on raw hard-disk tracks.

I store phone numbers in an array in non-volatile memory on my cell-phone.

One might even convince me that a place exists in this world for a simple network model object cache like Carl sells. Just don't try to tell me it's a database management system. It's not a system. It apparently provides no management facilities. And don't try to tell me it has any advantages over an RDBMS -- it doesn't.

> > > (Also, while Carl is reading about relational algebra and tuple
 calculus,
> > > Bob needs to read the series of responses to the Third Manifesto ...)
> >
> > I haven't seen those. Where are they?
> >
>
> Wow, a C.J. Date fan who doesn't know about his detractors? :)

Actually, I am quite familiar with many of his detractors, and I have read much of the object database literature. I just don't know of any works that directly criticize the Third Manifesto in any convincing manner.

Given that hundreds of thousands of pages are written on the subject of databases every year, please forgive me if I miss one or two.

> You should be able to find the articles on www.acm.org in the digital
> library. There was a series of exchanges in SIGMOD after the Third
> Manifesto was published.

If they were in the Record, I already read them. Please forgive me if I found none of them in the least convincing.

If you think my understanding of the issues is in any way lacking due to an article I missed, please give a specific reference or bring up the specific issue. I will check it out or respond as appropriate.

> Now if you're uncomfortable with refereed
> exchanges, stick with Date/Pascal's Database Debunking site.

Not at all. I am quite comfortable with refereed exchanges. I'm just astounded how much gibberish sometimes makes it past the referees.

> (Sarcasm aside, I respect C.J. Date ... I present his views for balance in
> my object data management course)

If you present his views solely for balance, surely your course addresses his criticisms of network model oodb. How does your course respond to the need for physical independence? To the need for logical independence? To the deficiencies of forced navigation? To the uncompensated additional complexity of network model OODBMS? To the simple observation that "object class" and "relational domain" are exactly the same thing: user defined type? To the myriad research regarding optimization, normalization, minimal constraint checks, view updatability etc. that are all based on the mathematical foundation of relations?

Just curious,
Bob Received on Sun Jul 22 2001 - 01:29:45 CEST