On Mon, 16 Apr 2001, Bip Thelin wrote:
> Kief Morris wrote:
> >
> > [...]
> >
> > I think this is a good way to go about it: it looks like the table name can be
>configured
> > in the server.xml file. Probably the column names should also maintained as
>JDBCStore
> > properties for configurability.
>
> Yes, that's why I wrote the last email to get some feedback on fields needed for the
>table.
> I'll implement it so it'll be configurable.
>
> > The table will need a column for the lastAccessedTime. And the ID column will need
>to
> > be something like CHAR, VARCHAR, or BYTE rather than an int.
>
> Yes, a typo from me. I looked at your latest patches for Persistentmanager and the
>Managerbase
> and it looks good. However we should come up with some smart solution when working
>with JDBCStore,
> the processexpires() in PersistentManagerBase is looking for current sessions then
>swaping them
> in and checking if they're valid, if so continue else invalidate. This is fine for
>FileStore but
> with JDBCStore this causes each session to be retrieved over the network(or where
>the RDBMS is located)
> just to check if the session is valid. Maybe we could add a method to the Store
>interface, i.e.
> public boolean isSessionValid(String sessionId) {}.
>
> What you would gain from this is that each Store is responsible for checking if
>expired in
> the way that's best for that particular Store, with JDBC you could do that in a
>single select
> statement without having to retrieve the data from the session(Assuming you save the
>lastaccesstime in
> the database, I'm currently working on that).
>
> ..bip
>
One of my original thoughts was along this line ... a Store should be
responsible for expiring its own swapped-out sessions. In practice, you
would have a background thread inside JDBCStore doing this for you. The
Store would also double check the current status while processing a
load().
Craig
