Richard S. Hall wrote:
> Hey guys,
> 
> I have changed the framework launching code in Main to try to eliminate
> some need for configuration. The idea is to have the launcher work a
> little bit like File Install in that it automatically will install and
> start any bundles found in the "bundle" directory. No existing
> functionality has been removed and for the most part it should look very
> similar to the previous behavior.
> 
> The benefit here is we no longer need to configure the felix.auto.start
> property to auto deploy the bundles in the "bundle" directory. If you
> want to add/remove bundles, you just add/remove them to/from the
> "bundle" directory. This is intended to be much simpler than File
> Install and only takes effect at framework startup with no monitoring of
> the directory at run time. (I originally thought about using File
> Install, but after a spate of threading issues with File Install in
> JIRA, I felt it was maybe too much.)
> 
> Currently, the only real noticeable change is that we now use the
> absolute path as the location of the installed bundle, rather than a
> relative path. This could cause some exceptions for existing caches, but
> they will still function. (We had some issues with File Install and
> relative URLs previously, so this change will work better with that in
> the future I think.)
> 
> One issue I was wondering about is whether or not I should make the
> launcher perform an update on bundles in the "bundle" directory when
> they are already installed. This would be a change from previous
> behavior, which always just installed so you would only see the existing
> installed bundle even if you changed the target. It seems like this
> could be convenient, but not strictly necessary. What do you think?
> 
> I do not think doing an uninstall of missing bundles would be a good
> idea, though. The whole point is to provide a simple, configuration-less
> way to get bundles into your framework at startup, i.e., a replacement
> for felix.auto.start. I don't want to create a complex provisioning
> solution...that is a separate project.
> 
> I'd like to get this wrapped up quickly, so any input is welcome. Thanks.
> 
I think it's great the way it is right now - it's sufficient for this
use case and if people need more for provisioning there are other
solutions available.

Carsten

-- 
Carsten Ziegeler
[email protected]

Reply via email to