IMHO this sounds like a rather good idea.  Defining a service using an
interface is unquestionably a best practice, but I'm not sure if it
should be imposed by the framework.  Especially when prototyping I think
it would be nice to be able to define services through their
implementing class only.

--knut

From: Achim Huegen [mailto:[EMAIL PROTECTED] 
> 
> Do you mean the PrototypeBeanFactory?
> This is an excellent proposal. By the way, it's mine ;-)
> Did you have time to look at the implementation I filed to jira?
> 
> Nevertheless its use is of quite different nature.
> I want singletons and the possibility to reference my pojo services
> from other services and configurations.
> 
> Achim Huegen
> 
> Howard Lewis Ship wrote:
> > You might want to look at BeanFactory.  A BeanFactory vends out
> > objects instantiated by factories, but it is much looser (and less
> > powerful) than services. There's also a proposal to have resources,
> > which are objects built by something similar to the BuilderFactory,
> > but not be services, just objects.
> > 
> > On Mon, 12 Jul 2004 22:04:27 -0400, James Carman
> > <[EMAIL PROTECTED]> wrote:
> > 
> >>Actually, if you use CGLIB (I don't know about the capabilities of
> >>javassist, as I'm no expert, so it might actually work 
> there too), you can
> >>create interceptors on POJOs too!  Spring already does this.
> >>
> >>
> >>
> >>-----Original Message-----
> >>From: Achim Huegen [mailto:[EMAIL PROTECTED]
> >>Sent: Monday, July 12, 2004 4:25 PM
> >>To: [email protected]
> >>Subject: POJOs as services
> >>
> >>I must admit I'm lazy, sometimes. Whenever I complete a new service
> >>implementation I think 'oh no, now I have to split my service into
> >>interface and implementation, just to be able to use it in hivemind.
> >>Dealing with two units which both contain the same methods 
> and use linked
> >>javadoc is extra work. Some service interfaces are not very 
> stable in the
> >>beginning of development and some services will never 
> require the use of
> >>interceptors or benefit from multiple implementations.
> >>
> >>So, why don't we support POJO (interface free) services?
> >>- It would ease adoption of hivemind for new users, by 
> reducing some of
> >>the overhead of using hivemind.
> >>- Integration of existing classes would be able without adapters or
> >>wrappers. For example you can setup a global 
> SimpleDateFormat instance
> >>without problem.
> >>
> >>Of course it would not be possible to use interceptors or 
> any service
> >>model besides primitive.
> >>
> >>In fact it is very easy to implement, just remove the 
> interface check from
> >>ServicePointImpl.lookupServiceInterface and hivemind is 
> ready for POJOs.
> >>Better: Let the serviceModel decide whether it want to support
> >>non-interface classes.
> >>
> >>Achim Huegen
> >>
> >>------------------------------------------------------------
> ---------
> >>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]
> >>
> >>
> > 
> > 
> > 
> 
> 
> ---------------------------------------------------------------------
> 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