Re: Dreaming About Redesigning SQL

Quoth "Anthony W. Youngman" <thewolery@nospam.demon.co.uk>:
> In article <3f8cbee1.1656673@shawnews>, Gene Wirchenko
> <genew@mail.ocis.net> writes
>>seunosewa@inaira.com (Seun Osewa) wrote:
>>
>>[snip]
>>
>>>Sometimes I wonder why its so important to model data in the "rela-
>>>tional way", to think of data in form of sets of tuples rather than
>>>tables or lists or whatever.  I mean, though its elegant and based
>>>on mathematical principles I would like to know why its the _right_
>>>model to follow in designing a DBMS (or database).  The way my mind
>>>sees it, should we not rather be interested in what works?
>>
>>     How do you know it works?  Without the theory and model, you
>>really do not.
>>
> And don't other databases have both theory and model?
>
> It's just that all the academics have been brainwashed into thinking
> this is true only for relational, so that's what they teach to
> everyone else, and the end result is that all research is ploughed
> into a model that may be (I didn't say "is") bankrupt. Just like the
> academics were brainwashed into thinking that microkernels were the
> be-all and end-all - until Linus showed them by practical example
> that they were all idiots :-)

In mathematics as well as in the analysis of computer algorithms, it
is typical for someone who is trying to explain something new to try
to do so in terms that allow the gentle reader to do as direct a
comparison as possible between the things with which they are familiar
(e.g. - in this case, relational database theory) and the things with
which they are perhaps NOT familiar (e.g. - in this case, MV
databases).

Nobody seems to have been prepared to explain the MV model in adequate
theoretical terms as to allow the gentle readers to compare the theory
behind it with the other theories out there.

I'm afraid that does not reflect very well on either those lauding MV
or those trashing it.  

- Those lauding it have not made an attempt to show why the theory behind it would support it being preferable to the
othermodels around.
 
 I hear some vague "Oh, it's not about models; it's about language" which doesn't get to the heart of anything.

- And all we get from Bob Badour are dismissive sound-bites that _don't_ explain why he should be taken seriously.
Indeed,the sharper and shorter he gets, the less credible that gets.
 
 There are no pointers to "Michael Stonebraker on Why Pick Is Not My Favorite Database."  Brian Kernighan felt the
issueswith Pascal were important enough that he wrote a nice, approachable paper that quite cogently describes the
problemswith Standard Pascal. <http://www.lysator.liu.se/c/bwk-on-pascal.html>  He nicely summarizes it with 9 points
thatfit on a sheet of paper.
 
 If Bob wanted people to take him really seriously about this, and has done all the research to back up the points that
areapparently so obvious to him, then it should surely be _easy_ to write up "Nine Reasons Pick Isn't My Favorite
DatabaseSystem."
 
 And just as people have been pointing back to Kernighan's paper on Pascal for over 20 years, folks could point back to
the"Pick" essay.
 

But apparently it is much too difficult for anyone to present any
_useful_ discourse on it.
-- 
(reverse (concatenate 'string "ac.notelrac.teneerf" "@" "454aa"))
http://cbbrowne.com/info/nondbms.html
For a good prime call:   391581 * 2^216193 - 1 
-- smr2@cornell.edu (Szymon Rusinkiewicz)