ObjectWeb Consortium
Search ObjectWeb Mail Archive: 

Advanced Search - Powered by Google


Mail Archive Home | proactive List | Febuary 2004 Index

<--  Date Index  --> <--  Thread Index  -->

Re: [fractal] comments on the Fractal 2.0 specification draft


Hello, 

thank you Eric and Jean-Bernard for your answers.

Here are some comments regarding your comments on our comments ...


Concerning collective interfaces :

As a matter of fact, our implementation also lazily creates the collective 
interfaces. I was just considering the case where the number of interfaces 
would not be infinite, but a fixed number.

Besides, in our implementation, the collective interface is materialized by a 
typed group of interfaces[1]. As such the user only sees 1 interface, but 
this interface has a collective behavior : the calls are either scattered or 
broadcasted to all connected server interfaces.[2]
It still is compliant with the spec (6.1), because the name of the interface 
still *begins* with the name specified in the type definition.


Concerning the convergence of Java implementations :

> On your implementation for Proactive, it would be nice if we could see how
> best to make your implementation and Julia converge. My sentiment is that
> we should have one toolkit for supporting Fractal components in Java and
> that if specific needs appear, we should be able to acomodate them as a
> proper extension to the toolkit. What do you think ?
>

I am not sure I understand your question correctly.
Are you suggesting using Julia as the unique framework for building Fractal 
component in Java?

We first tried to use Julia to manipulate active objects as components, but 
as 
such, we wouldn't have been able to reuse the features of the ProActive 
library. So we had to go for our own implementation of Fractal [3].

Our implementation is different from Julia both in its objectives and in the 
implementation techniques.

Our objective is to provide (or at least to investigate) the component 
paradigm for the development of distributed P2P and Grid applications.

The implementation is based on a meta-object protocol, the user classes are 
not modified during runtime, and the components benefit from all features of 
the ProActive library (asynchronism, distribution, mobility, activity etc.). 
An application based on such components is fully distributed, there is no 
such 
thing as a Julia runtime.
We currently provide specific controllers (LifeCycle, Binding etc) for the 
management of the components, but we could envisage facilities for adding 
custom controllers for satisfying specific needs, like in Julia.


Still, the api is the same, and we will be investigating interoperability 
between Julia and Proactive (notably the possibility of aggregating 
components from different implementations).

Convergence can also be achieved on the tools around the implementations. For 
example, with the Fractal GUI : we are trying to figure out a way to be able 
to reuse it as a conception / visualization tool, and to extend it towards a 
deployment tool. (I think a new release of the Fractal GUI should come up 
after the new ADL is clarified?)
The ADL is another point of convergence, where we had specific needs 
(deployment abstractions) ; these needs have be taken into account in the new 
version of the ADL.



The first version of the implementation that we are writing will be available 
around march or april with a new release of the ProActive library.



Cheers,


Matthieu


[1]http://www-sop.inria.fr/oasis/proactive/doc/api/org/objectweb/proactive/doc-files/TypedGroupCommunication.html
[2] By the way, this possibility is not mentioned in the specification.
[3] after discussing this in an "Arcad" meeting last year in grenoble with 
Jean-Bernard, Eric and Thierry






<--  Date Index  --> <--  Thread Index  -->

Reply via email to:

Powered by MHonArc.

Copyright © 1999-2005, ObjectWeb Consortium | contact | webmaster.