Thanks for your reply! On 03.06.2011 14:34, BJ Hargrave wrote:
> For a factory component instance created (and activated) with > newInstance, it does not really matter whether the instance is > activated before it is registered as a service or not. I don't understand why it doesn't matter. If the instance is registered first, then bundles can retrieve a not yet activated component instance via the service registry. Since activation includes running the activate method which may perform important initialization tasks, I think it should be required to activate the component instance first. > The main reason for (non-factory) components to be registered before > being activated is to allow for lazy loading of the component class > (and dependent classes). With a factory component, there is no need > for this since the call to newInstance must immediately load the > component class. Okay, I understand this point, thank you. > Since there is no opportunity for laziness, activation before > registration makes sense. Yes, I agree it makes sense, but I would go even further and think the spec must require it for the reason stated above. But even if I miss an important point here, the spec should IMHO do one of three things: (1) require activation before registration (2) require registration before activation (3) let the order explicitly undefined As I wrote in my first mail, the spec seems to do (1) at some places and (2) at some other places. Regards, Daniel
signature.asc
Description: OpenPGP digital signature
_______________________________________________ OSGi Developer Mail List [email protected] https://mail.osgi.org/mailman/listinfo/osgi-dev
