Re: [rgui-dev] Form vs. Function

[Massimiliano Mirra]

On Fri, Sep 06, 2002 at 06:27:01AM -0600, Tom Sawyer wrote:
> just had a notion about Form vs Function.
> 
> you were talking about not being a typesetter, so to speak, as applied
> to GUI creation, correct? 

Yes, though I admit I'm swimming in total confusion about the subject.
I tried to jot down something, and I found myself floating to higher
and higher logical levels.  But the farther I got from implementation,
the more I approached what might be an appropriate metaphor (and who
knows the way back to implementation might not be too long).

We've been talking about abstracting interfaces and writing markup for
their structure as if they were texts.  But text is a one way
communication.  What we've got here, instead, is a conversation.

We have speakers (the user and the computer), meanings (``I want you
to put the value 0.3 into the A2 cell'') and a (visual) language,
where language means a set of relations between some meanings and some
sensory messages.

Now you see what I meant by higher logical levels. :-) One question
I've been trying to answer is: is this too high a level to be useful
in practice?  The other is: how do I model a conversation that happens
in a visual language?

> i was wondering to what extent we could build
> gui-intelligence into GUtopIa for automatic GUI creation?
> 
> at the absolute highest level we would want to be able to feed GUtopIa
> the logic-model (core-app) and presto! out comes a fully made GUI. of
> course, this level of intelligence is about impossible. it wouldn't even
> be reasonable for a human to do, outside the core-app's developer. 

Absolutely.  The `typesetting' equivalent would be to expect the
program to understand umarked text and arrange it.

> so some hints, at least, need to be specified.
> 
> and so, that's kind of the notion i'm having. granted that without any
> hints whatesoever, GUtopIa will do a really crapy job of automatically
> generating a GUI, but that's okay becasue the more hints we give it, the
> closer it will get to a better design. are you following me here?

Yes.  I don't see it as a refining process, though.

function (or whatever) markup
                 |
                 V translation rules
                 |
presentation markup or other means of expressing presentation

In this context, `giving more hints' might mean a) refining the rules,
b) making the function description more exact, or c) manually
modifying the presentation, like you could with the HTML that is
generated from an XML document (function) processed with an XSL
stylesheet (rules).

> in other words, what if we created a gui-template markup, such that it
> doesn't require you to be fully explicit, and GUtopIa fills in the
> holes. of course you *could* define every last detail in order to
> completely control the layout if you wanted. the beauty of this is that
> one can then be as much or as little of a "gui-typesetter" as required
> to produce useful results. and of course the better we made GUtopIa's
> "Hueristc-Infrence Engine" the less one would need to be the typesetter.
> 
> of course, don't ask me how we'd implement such a thing. at this point i
> have no idea.
> 
> thoughts?

Just one: concentrate on the presentation layer (Chameleon?) to get
something usable as soon as possible that fulfills the base
requirements (Ruby way API, cross platform with native look and feel).
It's ok if it is used `low level' at first, just like any other
toolkit.  We can keep the brainstorm going about the upper layer, just
don't wait until it's done to write code as it won't be a short wait.


Massimiliano