> From: Stephen McConnell [mailto:[EMAIL PROTECTED]] 
> >2) Most agree that we should consider using Commons Logging 
> instead of
> >   maintaining our own.  Any dissenters I missed?
> >
> 
> Can I hold onto an option to dissent?
> I love log kit configurability and structure from a 
> management point of 
> view.  Provindg changes don't eliminate any of that 
> functionality I'll 
> be happy.  Otherwise you will be pushing me to dissent. :-)

Commons Logging is just another logging abstraction--developed at
the same time as Avalon logging abstraction.  Ours was released
roughly a month before theirs.


> >5) CM should at least look like this:
> >
> >interface ComponentLocater /* or ServiceLocator */
> >{
> >    Object lookup(String role) throws ComponentException;
> >    boolean hasComponent(String role);
> >}
> >
> 
> Which is basically equivalent to A4 ServiceManager assuming 
> release() is 
> depricated.  As far as A4 is concerned I thinkk we should 
> introduce this 
> into the service package (renamed to ServiceLocator for package 
> consistency), update ServiceManager to extend from ServiceLocator for 
> backward compatability, and declare ServiceManager as a depricated 
> interface.  That provides a complete migration path now - now need to 
> wait for A5.


:)


> >class ComponentException extends Exception
> >{
> >   // ... skip constructors and impls.
> >
> >   String getRole();
> >   Throwable getCause();
> >}
> >
> 
> +1
> Could be added to the current A4 defintion of ComponentException and 
> ServiceException without brreaking anything.

Yep.


> >All dissent is on anything extra.
> >
> 
> I think the main thing missing in the above is the work on getting a 
> formal meta model in place as part of the framework.  This 
> can be viewed 
> as an A4 activitiy because there are not obsticles and no 
> changes needed 
> to put it in place.
> 
> One final note - given the above - I'm really questioning the 
> necessity 
> for an A5 any time soon (blame that on Pete Royal for putting the 
> thought into my head, and yourself for presenting the above summary 
> which seems to negate a requirement for package name changes).  


The package name changes are mostly for easing the typing requirements.


> If instead if I look at the hot topics - Avalon 4.2
> 
>   - doc updates
>   - putting a locator in the framework.service package
>   - update the ComponentException/ServiceException to hold a 
> role reference
>   - focused activities on framework metadata
>   - focussed activites on common logging and how that relates 
> to logkit
>   - focussed activites on common container services 
> (assembly, security 
> policy, etc.)
>   - other commons related stuff to ensure communication of framework 
> resources
>     such as configuration
>   - component demos
>   - more demo
>  
> Then I get everything I wanted from 5.0.  
> ECM has a migration path via the service package.  
> Nothing breaks.  

ECM gets deprecated and Fortress lives on.  ECM is still a beast
worth killing.  It mixes Container/LookupMechanism concerns.  It
is not really extensible.

Fortress needs some attention, but we want to make it as extensible
as we can--without damaging maintainability.


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to