+1 It also seems a lot more intuitive to me to have service.ranking and filter.order both work in the same ordinality.
On 6/18/13 8:20 AM, "Bertrand Delacretaz" <[email protected]> wrote: >On Tue, Jun 18, 2013 at 2:04 PM, Felix Meschberger <[email protected]> >wrote: >> ...We basically have two options: >> >> (1) Keep the implementation and fix the documentation. This would allow >>us to keep >> the implementation and maintain backwards compatibility at the expense >>of not following the OSGi spec >> with respect to the service.ranking property... > >I'm in favor of this option, including writing integration tests that >demonstrate it (yes I volunteer ;-) > >I don't think the OSGi spec is a problem, we are ordering the services >based on that, but then you could argue that filter 1 should be called >first because it's 1, or that filter 123456789 should be called first >because it has a higher ranking. > >Let's not break backwards compatibility based on this arbitrary choice >of ordering. > >-Bertrand >
