RE: [Sweet-swt-develop] Has anyone here seen this?

[Dave O. <Da...@AS...>]

Nice work, Joe.  As you said, please do keep us informed.
 
Dave

-----Original Message-----
From: Joe Winchester [mailto:WIN...@uk...] 
Sent: Tuesday, June 17, 2003 6:46 AM
To: swe...@li...
Subject: Re: [Sweet-swt-develop] Has anyone here seen this?



Hi Dave, 

I spoke to Mark Davidson at Sun ( the developer looking after
JavaBeans/BeanInfo spec ) and we talked about having a whole ramp on the
BeanInfo spec.  Things we discussed were 

a)        Making it all go to XML because BeanInfo classes are really
declarative and not behavioral and for some folks ( those who are using
JavaBeans as objects they wrapper as web services and want to created WSDL )
it makes more sense. 

b)        Looking at metadata ( the URL you gave for this ).   The good
thing about metadata is that it is part of the source of the JavaBean ( so
you don't need a separate BeanInfo class or XML document to manage and
deploy ), and unlike comments that you can't parse and don't go into the
.class file, metadata does.   The disadvantage is that because it's part of
the class itself you don't get the advantage that BeanInfo does right now (
whether in the existing .class or an XML format ) that you can augment it
with your own.  For example, Sun provide their own BeanInfo classes however
we at IBM have our own ones for VisualAge for Java and WebSphere Studio.
This is because some of the information in the BeanInfo is put their by the
author and is really just declarative, however some isn't.  Things like
property editors are the "value add" that IDEs provide.  It did make me
think ! about this whole issue and how people currently extend BeanInfo by
having full replacement classes and slamming it into the Introspector's
search path.  We need a better way to just delta your metadata on top of the
compoent author's existing definition. 

There is a JSR that is just about to begin to ramp the whole BeanInfo spec.
Marc ( from Sun ) and I are pushing it through right now to get JCP approval
and given that IBM and Sun are backing it I doubt it'll get voted down.
Once this gets going then our thoughts are something along the lines of
metadata for the class author, XML for deltas, and allowing extension to
other component models not supported by BeanInfo including modelling
relationships ( container to component for example ) and also non-default
construction and other patterns.  We'll need an expert group once this gets
going so hopefully there is synergy between Sweet and the BeanInfo ramp.
I'm going to chase the IBM lawyers today to see where it's at in the JCP and
I'll post here once I know more. 

Best regards, 

Joe 


Please respond to swe...@li... 


Sent by:        swe...@li... 


To:        "'swe...@li...'"
<swe...@li...> 
cc:         
Subject:        [Sweet-swt-develop] Has anyone here seen this? 


 <http://www.jcp.org/en/jsr/detail?id=175>
http://www.jcp.org/en/jsr/detail?id=175 


Thoughts/opinions? 



Dave