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

[Joe W. <WIN...@uk...>]

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 
Thoughts/opinions? 

Dave