Okay, where should I package the code? Under org.jboss.jdbc or
org.jboss.ejb.plugins.jdbc? I'd suggest we create separate sub-packages
for:
- JDBC Pools
- 1.1 XA wrapper
- Configuration interface (this may go in the main jdbc package)
Only that last would be a "public" interface - the rest would be
accessible via JNDI or JDBC URLs only, or used under the covers.
Aaron
On Mon, 22 May 2000, Rickard [iso-8859-1] �berg wrote:
> Hi!
>
> Aaron Mulder wrote:
> > On Mon, 22 May 2000, Rickard [iso-8859-1] �berg wrote:
> > > Hold on!!! You have used a *connection* (a pipe, something to send stuff
> > > through, but *not* a data storage) to execute queries. All that data has
> > > been sent to the database. The data is *not* tied to the connection as
> > > such, it is tied to the *tx*. The connection is just a pipe we use to
> > > shuffle our stuff through.
> >
> > I wasn't aware you could distinguish between a connection and a
> > transaction in JDBC. But perhaps this is what you meant when you talked
> > about 2.0 drivers.
>
> Exactly. In JDBC 1.0 con's and tx's are tied very strongly to each
> other. Now, with the XAResource thing this coupling is much more loose.
>
> > So you're saying that with a "2.0" driver, the DBMS
> > vendor implements XAResources and XAConnections within their driver, so
> > we don't deal directly with the connections or transactions - they handle
> > it all for us. Magical!
>
> Right on :-)
>
> > I went back and looked at the Oracle drivers, and discovered that
> > that latest drivers do indeed support this - they provide an XADataSource.
> > That sures saves a bundle of work! NOW we're on the same page!
>
> Yep, that saves a LOT of work :-)
>
> > However, this simply raises a new and different set of questions.
> > Sigh.
> >
> > 1) How many drivers implement this? I suspect not too many, but I'd love
> > ot be surprised.
>
> Exactly, which is why I proposed a JDBC1.0 driver wrapper that makes any
> non-2.0 driver appear as though it was 2.0. Of course, it will always
> fail on "prepare" and "commit" (since JDBC 1.0 is 1-phase only!), so the
> tx-manager will have to make sure to use the 1-phase optimization if
> possible.
>
> > 2) Unfortunately, the DataSource interface does not include any methods
> > for configuring the DataSource. That is, there's no standard way to
> > specify the server, port, instance/database, URL, etc. The spec suggests
> > that the implementation be beany (get/set for all properties), but only
> > "suggests" the properties. So how do we allow configuration? I suppose
> > we can take a property list and try to invoke an appropriate setXXX method
> > for each proparty via reflection, and error out if we can't find an
> > appropriate method...
>
> Exactly.
>
> > 3) What do we do with drivers that don't implement this? I think we may
> > have the same intention here, which is to create a wrapper for 1.0 drivers
> > where we implement all the XA stuff and prevent bad things from happening.
>
> See above. Yes. Non-trivial, but the only way to go IMHO.
>
> /Rickard
>
> --
> Rickard �berg
>
> @home: +46 13 177937
> Email: [EMAIL PROTECTED]
> http://www.telkel.com
> http://www.jboss.org
> http://www.dreambean.com
>
