Re: data management

From: paul c <>
Date: Tue, 06 Jun 2006 01:49:36 GMT
Message-ID: <Qa5hg.243202$P01.19605_at_pd7tw3no>

David Cressey wrote:
> "mAsterdam" <> wrote in message
> news:44845403$0$31639$

>> * changed subject line *
>> x wrote:
>>> David Cressey wrote:
>>>> mAsterdam wrote:
>>>>> Patrick May wrote:
>>>>>>     I did not find him backing up his claims regarding Mr. Martin's
>>>>>> "foolishness".  Can you or can you not point out a specific example of
>>>>>> him doing so?
>>>>>> Robert Martin ha escrito:
>>>>>>>> Ridiculous. OO and RDB coexist very nicely together.
>>>>>>>> I've never heard anyone suggest
>>>>>>>> that searches aren't needed.
>>>>>> Relational databases have nothing to do with searches.
>>>>>> This shows profound ignorance about data management theory.
>>>> I disagree with both points.
>>>> The need to support searches is one reason for building a database in

> the
>>>> first place.  Perhaps the foremost reason.
>>> I would say relational databases are more
>>> about "finding" than about "searching" :-)
>>  >
>>>> The way relational databases support searches is one reason for choosing

> a
>>>> relational database over some other kind.  The same comment can be made
>>>> regarding SQL databases.
>>> The way relational databases support "finding" is one reason for

> choosing a
>>> database over a file system. :-)
>> The need to support searching/finding is a reason
>> for building fast access mechanisms, using indexes
>> (isam, vsam, ...) or otherwise.
>> DBMS's use these access mechanisms and provide an
>> easy-to-use interface to them. Anybody who has a
>> DBMS can simply use these as features.
>> The ease this brings may give valid economical reasons
>> to use a DBMS just to take care of searching.

> I think you're changing the subject. I suggested that supporting searches
> was a reason (perhaps THE reason) for building a database. You seem to be
> addressing the reason for building a DBMS. It isn't the same.
> BTW, enabling searches to find something is one way of supporting searches.
> In fact, searches that don't find anything can only be said to be productive
> if the fact that they didn't find anything carries some element of
> unexpectedness.
> This could bring us back to the curious incident of the dog in the night.

In the relational approach, what is a "search" except that it is nothing but a join? (on some attributes or none.) At a physical level, then, an index is one way to speed joins, but only one way. So I'd say that an rdbms requirement for searching doesn't necessarily imply a db requirement for indexes.

p Received on Tue Jun 06 2006 - 03:49:36 CEST

Original text of this message