Re: The proposed java policy have now moved.
[ Responding to old mail. The issue was whether Java packages
should depend on both java-virtual-machine and java1/2-runtime. ]
On Fri, Oct 05, 2001 at 05:18:29PM +0200, Ola Lundqvist wrote:
> > If all jvm packages out there specified whether they were java1-runtime
> > or java2-runtime compliant (or both), then all java packages could
> > depend on the runtime they require ?
>
> With this sheme it is possible to create a package that only provides
> the binary java (with a couple of other usefulthings) and an other package
> providing the core classes. I'm updating the policy right now because it
> want to fix the dependencies.
It's possible the other way, too. The idea is, Java applications
and libraries just depend on a runtime virtual package. Integrated
VM + runtime packages provide the runtime directly. Add-on runtimes
provide the runtime, and depend on java-virtual-machine, which can
presumably be provided by both standalone VM's, and VM + runtime
packages. Actually, this part may need to be more fine-grained: you
might need to specify the extension mechanism for the VM in the
dependency, eg java-virtual-machine-jni, java-virtual-machine-kni,
java-virtual-machine-cni. There might be other things I am missing
too, but I think the idea is right. It's definitely simpler and
cleaner for Java applications and libraries to depend on just one
virtual package.
Of course, you might need some more glue to tie this all together
(eg, making sure that the VM finds the appropriate runtime
libraries), but that is a separate issue (which the registry idea
tackles).
Andrew
Reply to: