On Thu, Apr 2, 2009 at 9:31 PM, Andy Schwartz
<andy.schwartz@oracle.com> wrote:
I was going to +1 - I like Kito's idea of using a component
author-specified attribute, and I also like Dan's idea of a more
generic "strict" attribute - though then it occurred to me that the
arbitrary attribute behavior isn't just a feature of Facelets
compositions - but this is also the way that Facelets works for any old
(Java-based) Faces component. That is, I believe that:
<h:commandButton foo="bar">
Just ends up stashing the "foo" attribute value in the
<h:commandButton>'s attribute map. (Or, at least, I think that's
what happens.)
Interesting, I never knew that, but it's true. I would assume that is how Facelets does the local variable assignments from attributes.
So now I am wondering whether there is merit in keeping
behavior consistent between composite and non-composite components in
this regard. Would it be confusing that arbitrary attributes are
pushed into the attribute map for some types of components, but not for
others? Hmm... maybe the benefits of providing quick feedback in cases
where the attribute is known to be invalid (ie. in the composite
component "strict" case) outweigh the drawbacks of the inconsistency?
I still don't really like the idea of using these random attribute names because it just destroys tooling. Am I being too rigid?
-Dan