My point is, if you use the extension you can't use a R5,R6 or R7
feature with that component.

Then we have things like FELIX-5775 where the code involved got so
complicated because of the proprietary support that I could find no way
to make this work when the extensions are used.

Regards

Carsten


David Jencks wrote
> As I recall, the behavior of all the extensions are available in the R6 
> version through the use of the ds ext annotations which get compiled into 
> Felix namespace xml attributes by bnd. What exactly are you proposing 
> removing?
> 
> I don’t think the top level xml namespaces offer much value but I’m not sure 
> the code supporting them is complex enough to worry about. For the most part 
> I think the behaviors of the extensions are fairly useful so I’d be reluctant 
> to remove them. Making an osgi server that uses them not able to upgrade to 
> r7 is a pretty big restriction.
> 
> David Jencks 
> 
> Sent from my iPhone
> 
>> On Feb 16, 2018, at 7:45 AM, Carsten Ziegeler <cziege...@apache.org> wrote:
>>
>> I would like to simplify the upcoming R7 implementation of Declarative
>> Services and drop the support for the proprietary extensions.
>>
>> Currently the implementation supports two additional namespaces for the
>> component XML. The oldest one allows to define an updated method before
>> it became part of the R4.2 release. So this has only been used
>> temporarily until that release became official.
>>
>> Similar, there is a special namespace for the R4.3 release which can be
>> used to indicate that the components might return additional service
>> properties from the activate method. This feature has been discussed at
>> that time in the OSGi expert group and was not rejected. While there
>> might be some people using this, this is tied to the 4.3 specification.
>> So as soon as you want to use R5 or R6 features or in the future R7
>> features, this proprietary extension is not available to your component.
>>
>> My suggestion is we drop these things for the R7 version and simplify
>> the implementation in some areas. If users want to use those features
>> the R6 implementation is still there to be used and we can maintain it.
>>
>> Regards
>> Carsten
>> -- 
>> Carsten Ziegeler
>> Adobe Research Switzerland
>> cziege...@apache.org
-- 
Carsten Ziegeler
Adobe Research Switzerland
cziege...@apache.org

Reply via email to