Hi all,
Still thinking about this implementation, I think I found a small
limitation in the current implementation. If I create a "code" for a
workspace name, and let's say a new workspace is created using the
Workspace.createWorkspace or Session.createWorkspace methods, I will
then associate a new code with the new workspace. Up to there all is
fine. Now it seems that Jackrabbit does not (yet?) manage the full
lifecycle of workspaces, or I haven't been able to find methods that
delete workspaces. The problem is that the PM should be notified of
this, otherwise there will be left-over data in the back-end. As far as
I could tell this problem would also arise with Filesystem-based PMs.
For the moment, if I understand this correctly, in the default
implementation and configuration, new workspaces will be created in the
repository/ directory. When using FS-based implementation, deleting a
workspace is trivial, you could simply shutdown Jackrabbit, delete the
workspace directory and all would be well (from what I read we might
also be able to "dispose" of a workspace, but I don't know if this is
correct) and then remove the files. Of course this is a bit manual and
not ideal for scenarios where the repository must have good uptimes.
In the case of a "disconnected" back-end such as a database, this
operation has to be done through the PM.
I should emphasis that I know this is not part of JCR, I'm only looking
at this from a Jackrabbit point of view. The other interim solution
would simply to let the data rot in the DB, in a disconnected state, but
that's not ideal either. The one nice thing about using different
database schemas mapped to workspaces is that disposing of a workspace
is easily done by a DB admin, but there is no SQL standard for managing
schemas.
Any corrections/input on this would be more than welcome !
Regards,
Serge Huber.
Edgar Poce wrote:
Hmm... I'm having another problem now. You suggested the use of a param
to set the name of the workspace. From the PM, how do I access this ? In
the PMContext class I only have accessors to things like the HomeDir,
FileSystem, NameSpaceRegistry, NodeTypeRegistry and RootNodeUUID. I
didn't see anything related to parameters. Has this been removed in the
current code ? Or are the settings set through setter and reflection API ?
It uses javabean convention, see
http://incubator.apache.org/jackrabbit/apidocs/index.html
BR,
edgar
Regards,
Serge Huber.
