Re: OO fans bashing Joins
Date: 2000/03/19
Message-ID: <25579fbc.5784a4f7_at_usw-ex0104-033.remarq.com>#1/1
In article <38d44a67.8577894_at_news.gte.net>, JRStern_at_gte.net
(JRStern) wrote:
>On Mon, 06 Mar 2000 06:02:41 GMT, topmind_at_technologist.com
wrote:
>>Is there some truth to this rumor, or is it just
>>more annoying OO propoganda?
>
>The later.
>
>Maybe relational is a bit dogmatic about not storing pointers,
which
>would not hurt anything as long as the full relational access
was not
>comprimised (the pointers could be used only when appropriate,
once
>upon a time there were such hybrid databases built and sold).
Of
>course it costs more at update time, to update whatever in-
pointers
>might exist, in order to increase performance at some expected
future
>select times -- often a good bet.
>
>RDBMS vendors and theorists have just decided, a priori, that
the
>performance penalty of recalculating joins is acceptable, and
is fully
>general, where non-relational pointer tangles would be fast at
what
>they do, but not fully general unless they are laid on top of
>something equivalent to a relational, normalized system.
>
>J.
>
>
I have heard a little about RAM-RDBMS. I assume they would try different indexing techniques than their disk-n-cache counterparts, such as using direct memory addressing.
As RAM drops in price, there should be more research or changes in this area I would suspect.
Thanks for your reply, -tmind-
- Sent from RemarQ http://www.remarq.com The Internet's Discussion Network * The fastest and easiest way to search and participate in Usenet - Free!