#: Marcel Reutegger changed the world a bit at a time by saying on 12/5/2005
11:54 AM :#
personally I think applications should only use a shared repository that
is deployed as a resource. either through some standard mechanism such
as RARs or some proprietary servlet engine mechanism.
e.g. use the jca contribution in jackrabbit.
In any case, the servlet engine or the application server should control
the life cycle of the repository (resource) and not the application
itself. (unless the repository is embedded into the application).
regards
marcel
I fully agree that normally the server should be responsible with this task. However, I don't
remember that the servlet engines are required to conform to JCA spec.
Doing more digging I guess an even easier way to accomplish what I have suggested and still not
expose the BindableRepository is to have the BindableRepository return true/false according to the
fact if it triggered the RepositoryImpl.shutdown(). The RegistryHelper.unregister method can check
this return and unbind the repository (the unbinding is still tricky as some of the JNDI are read-only).
./alex
--
.w( the_mindstorm )p.
Alexandru Popescu wrote:
Hi!
I have been reading about the decissions around the BindableRepository
and why or why not make it package vs public level.
I was wondering how to handle the following scenario:
2 applications A & B using a shared repository
application A starts the repository
application B retrieves it through JNDI
application A is shutdown for maintanance (the repository shouldn't be
closed, still being used by B)
JVM is shutting down
This results imo in a not-shutdown repository.
IMO a better approach would be to have the BindableRepository public and
its shutdown public, so that every application relying on it can `try´
to close the repository. I am saying `try´ because the
BindableRepository should be able to find out if it is still referenced
and trigger the RepositoryImpl.shutdown only when the last referencing
application is triggering the shutdown.
Please let me know what do you think about this. If you agree with this
scenario I can (almost) immediately provide the patched version.
thanks in advance,
./alex
--
.w( the_mindstorm )p.
!DSPAM:43940e7258489828750033!
