How about information such as excludes and aliases? We could display 
these in the method pane when the trait is selected in the category 
list, as you proposed when we talked.

This would also allow the editing. I think displaying this as virtual 
methods might be more natural, but I don't know how the editing would 
work then.

How about the fact that Traits at the moment have method categories? do 
we get rid of those? this would probably be ok for fine grained traits, 
might be a mess for large traits such as Andreas and Jecel were considering.

How about Traits that are composed of Traits? I think that information 
should be at least indicated, and probably also navigable.

Anyway, anyone feel like prototyping this based on the demo image? As 
long as you're working on that and modifying/extending the 
TraitsOmniBrowser package, the integration should be easy, and we might 
have both browsers in parallel for a trial period.

The code already in TraitsOmniBrowser should give some help implementing 
this - it implements a virtual category for the requires stuff, which 
might be a reasonable starting point for adding categories for Traits.

Daniel
PS - anyone working in that image, I do not at the moment advise 
upgrading the Traits packages on your own. We are still making 
occasional changes to the model that would break an image if you just 
load them. Just code your stuff so that it load into a new demo image, 
and we'll post new images from time to time until 3.9 merge time.


Avi Bryant wrote:
> On Sep 3, 2005, at 3:19 PM, Blake wrote:
> 
>>
>>
>> Ooh. That's a damn cool idea.
>>
>> Would that mean that Smalltalk categories would have (or reflect,  
>> rather) semantic value?
> 
> 
> We'd have to decide whether we wanted to just have traits (and so  every 
> method category would become a trait), or whether we wanted to,  at 
> least at first, have distinct concepts of first-class traits and  
> second-class categories.  I'd personally lean towards starting with  the 
> latter, slowly promoting individual categories into traits by  hand over 
> time, rather than a big-bang conversion of every method  category in the 
> image to a trait.
> 
> But yes, it would at least allow for some categories to have real  
> semantic value.
> 
> Avi
>