[JSR-77]

[Sean Hamblett <sh...@charter.net>]

I have seen a lot of messages on JSR-88, but haven't seen any recent messages 
on JSR-77.  I was reading the JSR-77 spec, and it seems to be based on the 
concept of Managed Object, and has an extensive object model, that I am 
having a hard time correlating to the existing codebase.  Is there an 
organized effort or plan to adopt the J2EE Managed Object Model, or has this 
topic not really come up on the radar yet?  I would like to participate in 
bringing it to life, but I would like a concensus on a direction/plan to 
start implementing this.  If anyone has been working on it, I would like to 
get some feedback on what should be implemented (all or parts), and an a 
tasklist for what is left to do.

thanks,

Sean

Re: [JSR-77]

[Jan Bartel <ja...@mortbay.com>]

Sean Hamblett wrote:
> I have seen a lot of messages on JSR-88, but haven't seen any recent messages 
> on JSR-77. 

There has been a thread on this topic running for a few days (indeed you 
have been one of the posters). See messags with a Subject line including 
"JSR77 component lifecycle". Here's a link to the mail archive: 
http://nagoya.apache.org/eyebrowse/SummarizeList?listId=140


  I was reading the JSR-77 spec, and it seems to be based on the
> concept of Managed Object, and has an extensive object model, that I am 
> having a hard time correlating to the existing codebase. 

There has been some work recently to bring the Component and Container 
objects in org.apache.geronimo.common more into line with the 
StateManagement methods and state transitions defined in JSR77.

> Is there an 
> organized effort or plan to adopt the J2EE Managed Object Model, or has this 
> topic not really come up on the radar yet?  

There are no JSR77 jmx classes checked into the repository yet, as the 
architecture of the geronimo services, the inter-service communication 
mechanism and the nature of support for JSR77 and jmx in general is 
still under discussion.

I would like to participate in
> bringing it to life, but I would like a concensus on a direction/plan to 
> start implementing this.  If anyone has been working on it, I would like to 
> get some feedback on what should be implemented (all or parts), and an a 
> tasklist for what is left to do.

I'm not sure the way forward is defined enough for a task list?

I think maybe the first task is to clarify exactly what the relationship 
is between the geronimo Component and Container and the concept of a 
"service", then move onto the JSR77 instrumentation of that.

ie is the relationship:

    service 1 -- 0..n container 1 -- 0..n component

Or is it:

    service 1 + -- 0.. n container 1 -- 0..n component
              |
              0..n component


At the very least, IMHO we need an AbstractContainer that models the 
1--n relationship with Components, such that a startRecursive() on the 
AbstractContainer will call start() (or maybe startRecursive() ???) on 
the Components.

As for actually starting the implementation of JSR77, then I guess as a 
starting point, we could implement the type hierarchy according to the 
JSR spec as plain Java interfaces. As a starting point Component should 
extend ManagedObjectand StateManageable, and AbstractContainer provide 
implementations of these methods.

Note that nothing needs to be jmx about this - the only place where jmx 
comes into play in JSR77 is the implementation of the MEJB. If we wish 
to have mbean representations of the ManagedObjects outside of the MEJB, 
then the suggestion has been made to use Dynamic MBeans to achieve this.


Jan

Re: [JSR-77]

[Greg Wilkins <gr...@mortbay.com>]

I have patched the Component and State classes so that they at
least conform to the JSR-77 state model.

There has been a patch submitted to add the StateManeable interface
to the code and I'm looking at that now.

Unfortunately other than the simple state stuff, I have not had
time to look at the rest of the JSR77 object model in detail, but
as I think our Container interface also needs some improvements, I
see no reason why not to be inspired by the JSR77 Object Model.

These may not end up being exactly the JSR77 MBean - but at least
if they are not, the mapping will be trivial.

cheers


Sean Hamblett wrote:
> I have seen a lot of messages on JSR-88, but haven't seen any recent messages 
> on JSR-77.  I was reading the JSR-77 spec, and it seems to be based on the 
> concept of Managed Object, and has an extensive object model, that I am 
> having a hard time correlating to the existing codebase.  Is there an 
> organized effort or plan to adopt the J2EE Managed Object Model, or has this 
> topic not really come up on the radar yet?  I would like to participate in 
> bringing it to life, but I would like a concensus on a direction/plan to 
> start implementing this.  If anyone has been working on it, I would like to 
> get some feedback on what should be implemented (all or parts), and an a 
> tasklist for what is left to do.
> 
> thanks,
> 
> Sean


-- 
Greg Wilkins<gr...@mortbay.com>             Phone/fax: +44 7092063462
Mort Bay Consulting Australia and UK.          http://www.mortbay.com