On Thu, 2004-03-11 at 18:51, Hamilton Verissimo de Oliveira (Engenharia
- SPO) wrote:
> -----Mensagem original-----
> De: Peter Courcoux [mailto:[EMAIL PROTECTED]
> 
> > It appears to me that the A4 framework implies a series of individual
> > contracts Configuration, Contextualisation, Servicable etc. AND an order
> > of execution.
> 
> Yup, although its sometimes implict.

But perhaps it shouldn't be...

> 
> I think I disagree with LS thoughs in sense of we are delivering an Avalon
> container, so - damn its Avalon - the container should know the steps of
> execution that he must obey, and guarantee its order. It must pass on the
> future tck. 

I agree that there should be an Avalon reference implementation and a
tck. I agree that the container should know the steps of execution that
it must implement AND their order.  This is a set of contracts and an
EXPLICIT statement of the order of execution.

> 
> Other strategies like Pico or Spring could be plugged into one of these
> execution steps - the avalon ones - not requiring a different handler, just
> different aspects/concerns implementations.
> 
This is just an implementation detail.


> The current situation of Aspects proposal is more like a generic component
> container where Avalon is just one of the possibilities. I think being a bit
> more restrictive wouldn't kill us :-)   but I might be wrong!
> 
But I think that you can have the best of both. Avalon 4  contracts that
specify particular concerns and the Avalon 4 specified order of
execution. 

But merely by changing the execution chain specification and adding or
removing the contracts you can move to Avalon 5, 6, 7 ... without
changing the container.

And if you are writing components for a j2me app you can lose the bits
you don't need and still be Avalon compliant. Your component will still
run in the reference implementation container. 

> About turning some phase on/off - at first this was my intention in castle,
> but in a second though I don't think its good. The user turning off a
> configuration phase for instance can guarantee that it won't affect his
> components, but can't say anything about 3th parties components being used.
> So, if the kernel don't have a concern/aspect specified for a particular
> phase it should use a default one.
> 
An Avalon 5 compliant container will have to provide the Avalon 5
facilities and the Avalon 5 specification chain of execution. Defaults
are fine. But why not allow the removal of unused bits by those
developers who don't need them? 

> just my 2 cents as usual.
> 
and my £0.02p ;-)
> 
Peter

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


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

Reply via email to