Re: [picocontainer-dev] MBean for pico?

Konstantin Priblouda wrote:

> 
> --- Paul Hammant <[email protected]> wrote:
> 
>> I do.  We started that before.
>> 
>> The idea of ComponentMonitors being swapple in a
>> running container
>> was that MX4J (or beanshell or grash) could
>> administer it, and
>> visualize things.
>> 
>> Also DefaultNanoContainer has named child containers
>> - again for
>> visual representation of container and component
>> trees
> 
> 
> Well,   now current version of nanosar rigs up pico
> container and places it onto JNDI. I think "start"
> container is also good idea.
> 
> But I lack a good idea how do I force JNDIExposed to
> actually bind - it has to happen right after component
> creation of course, and pretty early.
> ( say - start , then bind all exposed to JNDI possibly
> creating them if not created yet )
> 
> Shall I do this by visitor? And expose to JMX too?

Why don't you look at nano-remoting? That was already solved there (yes, by
a visitor). But also the stuff with the ObjectNames.

> Another Issue I got today is serialisability.
> JBoss likes to serialize everything that is bound
> to JNDI context.  And there is JNDIObjectReference,
> which of course holds JNDI-context - which is NOT
> serialisable. OK, I marked it transient for first -
> but this is somehow uncool. Does anyone have a good
> idea how to restore it after deserialisation?
> 
> "new InitialContext()" should be fine  in JB cluster,
> but I'm not sure about it.
> 
> just a random thoughts after forcing kids to bunk ;)

Expect an IC in the container used to resolve the component. Please, take
really a look at the old JMX stuff.

- Jörg


---------------------------------------------------------------------
To unsubscribe from this list please visit:

    http://xircles.codehaus.org/manage_email