On 19 May 00, at 18:01, Oleg Nitz wrote:
> Dan OConnor wrote:
> DO> Of course, SOMEHOW that reference must be matched up to an
> DO> actual bean. Agreed? Apart from an ejb-link or a proprietary
> DO> method, what is left? I can see only one other alternative from the
> DO> information in the deployment descriptor reference: the home and
> DO> remote classes. But this is not sufficient, because multiple beans
> DO> with the same home and remote classes may be deployed in the
> DO> same application (with differenent names and perhaps different
> DO> environment entries).
>
> DO> So what is left? My belief is that either the ejb-link mechanism
> DO> must be used, or the reference must be bound to a bean using a
> DO> proprietary method. I don't see any other way.
> What about ejb-ref-name?
> This is the name of the referenced bean that is known to the Bean
> Provider. I think that is most cases the Bean Provider knows the right
> ejb-name of the referenced bean. The ejb-link mechanism should be used
> by Application Assembler only if the ejb-ref-name is wrong
> (e.g., the referenced Bean was renamed, or it was obtained from
> other department).
>
> Best regards,
> Oleg
Hi Oleg,
I see what you are saying. Of course, if the recommendations of
the spec are followed, the ejb-ref-name will never match the
referenced bean name exactly... because one will be XXX, and one
will be ejb/XXX.
But we could have an option that trimmed the "ejb" off the name
and tried to find a matching bean. This is spec-compliant, because
it falls under the case of "proprietary deployment tool." :-)
This is a "reasonable defaults" case that I think we should support.
Great idea, as far as I'm concerned. So we would check (in the
following order):
1. ejb-link, because we are mandated to respect this information
at deployment time by the spec.
2. Any information in jBoss-specific deployment descriptors. (This
is Rickard's "proprietary method," and should be checked second
to give the bean deployer the opportunity to override the information
programmed in the bean.)
3. The ejb-ref-name, both trimmed and untrimmed (since putting it
in the "ejb" context is a recommendation, not a mandate, of the
spec.) I think we should call this the "Oleg default." :-)
Count me in, unless somene has a compelling argument against it.
-Dan
--
--------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Problems?: [EMAIL PROTECTED]
