Re: Minimal set of component attributes?

Berin Loritsch wrote:

> Stephen McConnell wrote:
>
>>
>>>
>>>
>>> This one.  It is much easier to mark an interface with a flag
>>> than to rely on the fact that I have a memory like a seive.
>>
>>
>>
>>
>> LOL :-)
>>
>> Know what you mean.  However, it's problamatic in that it locks you 
>> into a model where the interface can only be a work interface and 
>> cannot be applied as a internal interface.  I've thought about the 
>> same approach with respect to lifecycle stages in Merlin and reached 
>> the conclusion that its too restrictive - ended up focussing on 
>> explicity declaring the published services, lifecycle extensions, 
>> etc.  in the component. But keep in mind that what I have been doing 
>> is still real early.
>
>
>
>
> Who said you can't have it both ways?  I have seen nothing to restrict
> us to forcing only one way on the attribute declarations.


If you use the same tag to declare that an interface is a service, and 
to declare that a class is exposting a service, then you lose the 
ability to verify placement relative to context.  I agree that it would 
be feasible to have both, but they should be different statements - for 
example.:

  @avalon:service.defintion  // marks an interface as a service

  @avalon:service org.apache.playground.Demo // this component class 
exports the Demo interface

This enables you to verify that a tag is within a valid context - i.e. 
avalon:service.defintion would be valid in the header of an interface 
whereas avalon:service would be valid i the header of a class.

Cheers, Steve.

-- 

Stephen J. McConnell
mailto:[email protected]
http://www.osm.net