It looks like I have some pretty good idea of what needs to be done with
connection pooling.
Basically to allow for connection pooling with *any* versions of JDBC
drivers, also supporting newest advanced features, we need to create the
following things:
1. ConnectionPoolManager class - source of all database connections for
our DataNodes
2. class implementing javax.sql.ConnectionPoolDataSource interface - used
when a corresponding driver (or app server) does not provide one.
3. class implementing javax.sql.PooledConnection interface - used when a
corresponding driver (or app server) does not provide one.
4. class implementing java.sql.Connection interface - used when a
corresponding driver does not support pooled connections. Our
implementation will wrap existing connection object, delegating all the
calls to it, except for "close" method that will be implemented to return
connection to the pool.
In my opinion it covers all of the scenarios - DataSources and
PooledDataSources provided by app server (or by JDBC driver vendors) and
older JDBC drivers with no support for pooling. Basically we will be using
our code for each step (2. - 4.) only if a corresponding environment does
not have its own implementation. Only ConnectionPoolManager class is
always required since none of the Java specs define how pool should be
implemented.
Andrei
At 11:06 AM 4/2/2001 -0400, Andriy Shapochka wrote:
>They offer a connection pooling broker tested in the Oracle environment
>under GNU license. We can examine their code as we'll have to deal with
>all this connection stuff in a while
>
>http://www.javaexchange.com/
>
>--
>Andriy Shapochka
>
>--
This archive was generated by hypermail 2b30 : Sat Aug 04 2001 - 16:21:24 EDT