Re: Object-relational impedence

From: Marshall <>
Date: Tue, 18 Mar 2008 13:05:30 -0700 (PDT)
Message-ID: <>

On Mar 18, 9:13 am, S Perryman <> wrote:
> Their system used an RDBMS. And it performed poorly.
> The same systems subsequently built on the same platforms (HW, OS, comms,
> prog langs etc) , but using an OODBMS instead, performed orders of
> magnitude better.
> That's life.

I just want to drop a general comment in here.

Tools designed directly for a specific purpose will outperform general purpose tools.

If you have a narrowly-targeted domain and a domain-specific language that was designed for that domain, it will outperform a general purpose programming language in that domain. Take string matching for example, and regular expressions. Write a matcher for a complicated
pattern and put it in a high-performance matcher, and it'll kick the pants
off of anything ad hoc you care to write in any general purpose programming
language, both in brevity and speed.

If you have a specific function you want to execute in hardware, a custom-designed chip will beat your Pentium for that problem, with fewer gates.

If you have a data-management problem that you have a good understanding
of and is stable, and you care to take the time to write a special- purpose
data manager tailored just to that application, with query bias in favor
of the queries that you already know you're going to be running, etc., it will almost certainly be a lot faster than a general purpose DBMS, regardless of the data model of the DBMS.

Marshall Received on Tue Mar 18 2008 - 21:05:30 CET

Original text of this message