Re: PATCH: Bug 186

On Thursday, March 27, 2003, at 07:59  PM, Charles Goodwin wrote:
> Using your alternative method just shifts the problem elsewhere.  As 
> above.
> thisbox[2] != $s, so your box order has simply been compromised 
> elsewhere.

Oh yeah, I know that in practice it has very little difference with 
your model. However, the theory is the important point here. The 
developer is *actively reassigning* the location of the box, whereas in 
yours the pane simply moves it based on a little variable inside the 
box. It's easy to see in the example you gave, but if that variable is 
set in some other custom template accidentally, it could take a while 
to debug.

It's just a conceptual thing. Boxes should stay where the programmer 
left them. If that is followed 100%, it becomes a very useful rule that 
developers can simply rely on. Redirect and child wrappers are a 
transparent way of moving boxes without breaking the model. Or else, 
there really wouldn't be a point for such things to exist.

These little things are what all other graphics libraries seem to get 
wrong in my opinion. Every now and again, some basic design principle a 
library depends on changes, just for this instance (cough JTable).

d


_______________________________________________
http://lists.xwt.org/listinfo/dev