World: r3wp

Couldn't you do something like this, using compose?

actions: [

 my-text-list [set-face my-code-area read-string pick files value]
]

view compose/only [
	my-text-list: text-list files do (actions/my-text-list)
	my-code-area: code-area
]

And the same for every other possible event?

Sure ... but just looking at cleaner ways of doing this separation 
... getting rid of the 'do

It should be pretty easy to write a function that maps an ACTIONS 
block such as the following into a window:

	actions: [
		my-text-list [
			do [set-face ... ]
			"other-event" [ blah blah..]
		]
	]

where "other-event" could be on-click etc..

then the window doesn't have to specify any actions.

But I know what you're saying. You probably want such a function 
built in and used in the examples, so it becomes commonplace.

Yes

Maybe there could be a stylize type of way to deal with actions

Function name and arguments?
set-actions: func [window actions ...

anonymous windows?

I think it should be defined inside the window definition

set-actions view window-spec actions-block

(Assuming VIEW returns the newly created window object.)

modal windows??

(which it does.)

don't return until closed

AFAIK

Ok, so you want the window definition to separate internally into 
two parts.

Essentially:

view [
	; ACTIONS
	my-text-list/action: [set-face ...]

	; STRUCTURE
	my-text-list: text-list data files
]

yes just like a html page

(I'm still sort of using R2 terminology...)

and there's also passing the face correctly to the actions

Well, it should probably be in the reverse order, I suppose. Structure 
is put in place first, then behaviour operates on top of that.

view [
	; STRUCTURE
	my-text-list: text-list data files

	; ACTIONS
	my-text-list/action: [set-face ...]
]

makes it easier for the view layout engine I guess

STRUCTURE
 is actually "NAME-REFERENCE, STRUCTURE, CONTENT"

I'd also like to have it so that we don't have to name objects and 
can point to them based upon order ...

ahh... retract that.

I've spent some time thinking about that in the past, and it's difficult, 
if not impossible.

in case we wish to dynamically insert objects into the display and 
then remove them as they do in Javascript

yes, exactly. Dynamic structure.

Why?  If there's a tab order ... then there must be a serial structure

Which can be messed up all too easily by dynamic structure. Your 
program could reorder tabs for instance.

Or, as you say, a tab could be inserted/removed.

Eg.  I like to create tables first as instant feedback, and then 
fill them with data from an async function.

But I need to overlay a busy graphic which I can then remove ...

in the callback

Ok, so you as the programmer know that your table has a specific, 
static order at the time you need to fill it. No problem.

Did you also mean that you'd like ACTIONS associated with STRUCTURE 
elements anonymously?

view [
	; STRUCTURE
	text-list data files  ; <---- element #1
	code-area  ; <---------- element #2

	; ACTIONS
	[on-click [set-face ...] ] ; <---- actions for element #1
	[on-enter [ ... ] ] ; <---------- actions for element #2
]

well, I don't want to name each element of the structure

but maybe we have to .

We could combine both approaches. In the ACTIONS section, if there's 
a word followed by a block, then the word is the name reference to 
the element (eg. 'my-text-list), but if there is just a block, then 
the next element is taken anonymously.

Any objections to this vid enhancement?

not all structural elements will have an action

view [
	; STRUCTURE
	my-text-list: text-list data files ; <------ element #1
	code-area ; <--------- element #2

	; ACTIONS

 my-text-list [on-click [set-face ...] ] ; <----- actions for my-text-list

 [on-enter [ ... ] ] ; <------- actions for the next anonymous element 
 (in this case code-area).
]

I guess they will ... on-hover, on-tab etc

Yes, that could make some confusion. Specifying empty brackets [ 
] could act as a "ignore this element".

or, [] = don't override the default actions