Re: [ecf-dev] State of the 119 distribution part

[Scott Lewis <slewis@xxxxxxxxxxxxxxxxx>]

Hi Jan,

Rellermeyer Jan Simon wrote:
> <stuff deleted>
> > This brings up a spec question for me...what is the desired/spec'd
> > behavior when multiple distribution systems are present (e.g. cxf,
> >     
> ecf,
>   
> > etc)?  Are all distribution providers expected to respond?
> >     
>
> Yes. I am pretty sure that this is the case. 

OK, I've modified and released to HEAD changes to remove the code that 
only responds to "ecf" distribution service property...and removed your 
// JR comment.  So now our service hook will respond to all remote 
registrations.

Note...I've also updated the org.eclipse.ecf.osgi.services bundle 
contents (the osgi 4.2 compendium classes for discovery and 
distribution).  See CQ https://dev.eclipse.org/ipzilla/show_bug.cgi?id=2795.

A question...I see you marked the 
DistributionProviderImpl.getPublishedServices():

   public ServiceReference[] getPublishedServices() {
       // JR: XXX SMELL: this should not access the internals of the 
discovery
       // part since the discovery is optional.
       /*
        * return org.eclipse.ecf.internal.osgi.services.discovery.Activator
        * .getDefault().getServicePublicationHandler() 
.getPublishedServices();
        */
       return null;
   }

I started a new implementation of getPublishedServices().  I didn't 
complete it, because there is a question about desired behavior.  What 
if a (remoted) ServiceReference actually results in several publications 
(e.g. several ECF providers each register a ServicePublication for a 
given service reference?).  What should the 
DistributionProvider.getPublishedServices() actually return in that 
case?  It could return ServiceReferences associated with the 
*ServicePublications* (possibly several for each originating 
ServiceReference)...or it could return the ServiceReferences (of the 
remoted service) that have been published once *or more* (e.g. with >= 1 
publications)?  Any idea what the desired/expected behavior is here?

Scott