> From: Leo Sutic [mailto:[EMAIL PROTECTED]]
> 
> Yet the way such concensus based development would work may not 
> be ideal - in particular, the scope of the concensus may be too
> wide.
> 
> The problem, as I see it is this:
> 
>  + We need a very stable framework. Concensus here is vital.

It has been the model of what it means for consensus.  There are
a number of varying different oppinions.  I agree that Framework
needs to be something we all agree on.  That also means that
alot of really neat ideas don't become part of Framework.

Using a standard mechanism to extend the base Framework will help
in experimenting with ideas and proving their worth.  That way
we have imperical data on what it means to incorporate the new
contracts that a new lifecycle interface would entail.  We know
what it takes to implement a feature.  We have something now that
we can vote on whether it becomes part of Framework or not.

Until such a time, the best we can do is mental masturbation.
"It would be nice if...." and after going down the path you find
that there is no nice way to implement it--but you have to because
it is part of Framework....


>  + We can not develop a framework from first assumptions,
>    that is, the framework must be tried by users so we get
>    feedback.


Exactly.  Hense the common extension mechanism (it does not have
to follow what both Fortress and Merlin agreed upon).


>  + In order to have the product tested by users, we must have
>    a compelling product.

Correct.


>  + As users will only consider a product compelling if it solves
>    their needs, that product will have to evolve to meet those 
>    needs.

Correct.  But does that mean we have a reference implementation,
and then point others to a more robust solution (possibly hosted
by a non-apache location)?


>  + The rate of evolution for the product and the framework
>    will be different. The framework will only adopt the parts
>    of the product that "worked", and will thus evolve more 
>    slowly.

Yes, but the Framework should not define anything about the
container's implementation.  Just its contracts with the components.
We have to supply a compliance test suite.  That way we can have
an Avalon Test Compliance icon that can be used to endorse other
containers.  That gives users the confidence to try other containers
with the full confidence that they have something that will support
their application.


> Now, how do we handle those shear forces we have due to the
> different rates of evolution?

Same way we always have.  Containers are not part of the Framework,
but Framework needs to reflect not only the contracts of the
component to the container, but the container to the component.

> A question: what is the smallest unit that can claim concensus?
> Avalon does not need the consensus of Jakarta-Commons, we can
> vote ourselves. At the same time, I don't want one person
> creating a subproject of Avalon, declaring that he owns that
> code and that his own opinion is equal to concensus.

Check the three tiered proposal.


> In short, is there any way to decouple Phoenix from Avalon
> in such a way so that *all* PMC members need not vote on
> *every* change to Phoenix, yet keep Phoenix from sliding
> away to a little isolated group of its own?


PMC members do not vote on Phoenix technical direction.  Avalon
committers vote on Phoenix technical direction.  The fact that
some of the committers happen to be PMC members is only a
coincidence.

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

Reply via email to