Re: [stellation-res] Database abstraction

["Mark C. Chu-Carroll" <mcc@xxxxxxxxxxxxxx>]

On Tuesday 13 August 2002 07:55 pm, Rodolfo M. Raya wrote:
> Hi Mark,
>
> It seems to me that you are complicating things more than necessary.
>
> You are targeting 6 databases that need 6 different SQL scripts, not
> 6.000 different ones. The Postrgres version is ready; the Oracle version
> is working; don know the status of DB2 and I can modify the script for
> Firebird/InterBase. The only missing scripts are MySQL/InnoDB and you
> can search for a volunteer to fix the script when someone shows interest
> in using MySQL or InnoDB.

As I said in the note to Florin: it's not quite that simple.

It's true that we're ultimately looking at less than a dozen
databases. *But* we're building a system that's designed to
host extensions that all create and manipulate database tables.
Based on what I've seen talking to people, we could easily wind
up with 50 different extensions, all using the database. (There's
almost no chance that anyone would ever want to use all
50, but everyone will want *their* combination of 5 or 6.)

The problem that I'm trying to approach is making it possible
for extension authors to write the code to access and manipulate
data once; distribute thier extension as a binary jar; and then have
it work when it gets installed, regardless of the backend database.

So we're not talking about a solution for getting rid of the
multiple access points for different databases. That's an
unavoidable issue, due to differences in how connection
establishment and authentication works in different databases.
What we want is a solution that makes it easy to write new
code, new agents, new triggers, and other types of extensions
that manipulate databases in a portable way.

My original proposal is probably overblown. But I think
we need *some* way of papering over the differences between
databases, particularly when it comes to table creation. (Or maybe
only when it comes to table creation; I think that the only thing that
we've seen vary between databases that affects extension writers
is DDL.)

> Please e-mail me the latest Postgres  script  and I will give you
> tomorrow the Firebird version.

The problem is, there isn't a single postgres script. There's 
PostgresAccessPoint, and all of the *Agent classes.

	-Mark

-- 
Mark Craig Chu-Carroll,  IBM T.J. Watson Research Center  
*** The Stellation project: Advanced SCM for Collaboration
***		http://www.eclipse.org/stellation
*** Work Email: mcc@xxxxxxxxxxxxxx  ------- Personal Email: markcc@xxxxxxxxxxx