On Fri, Oct 3, 2008 at 12:07 PM, Ray Ryan <[EMAIL PROTECTED]> wrote:
> I'm with Isaac. I think the case for teaching our Widgets to implement
> HasData<T> is really clear cut (especially if they also accept DataChange
> listeners). The DataManager is a bit harder to justify, and anyway trivial
> for folks to implement on their own.

I'm not with Isaac.  My first experience working with data binding was
as a co-op student working on the UI team on the Eclipse project at
OTI.  That was a very educational co-op term!  Eclipse, at least in
version 2--I've lost track since then, has a very clear separation
between SWT, the widget kit, and JFace, the "data binding" library
that sits on top of SWT.

In my mind, a Widget has a specific job to do--get displayed in an
application and, when appropriate, react to user manipulation.
Teaching basic widgets to know about "data" is a bad idea, IMO,
because it makes basic widgets too complicated.  It also puts a lot of
pressure on programmers to use the data binding model that's baked
into the widgets, even if that model doesn't really suit a given
application.  In fact, I think Isaac's comments may be an example of
this--he's built himself a model for doing data binding and, at first
blush, it looks like it doesn't match with Arthur's.  It also seems
like neither Arthur's nor Isaac's model fits with the model that I
built and am frantically documenting in preparation for releasing
here.  If any of our models is chosen as The GWT Way and baked into
the basic widgets, the others are probably out of luck.  Instead, I
think a data binding library should be built on top of a widget
library for two main reasons: you can switch data binding libraries
more easily and, perhaps more importantly, you don't have to use any
data binding library at all.

Ian

--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---

Reply via email to