Re: Tile frame widget

From: Donal K. Fellows (donal.k.fellows_at_manchester.ac.uk)
Date: 12/21/04 

Date: Tue, 21 Dec 2004 12:08:59 +0100

R. T. Wurth wrote:
> I think this shows a defect in the entire TIP process, namely that
> it is too Tcl-developer centric. Did the TIP folks poll us, the
> real users about whether this new way of doing GUIs would be useful?
> Or about whether it would be sufficient to allow us to deliver
> software that meets our customer's needs?

At this stage, the Tile stuff has not reached the point of a serious
TIP. I anticipate that it will do, but it has not yet got that far.
Right now, Tile is just a way-cool extension to Tk that answers the
needs of a some commercial users of Tk for a L&F that is more like that
used by the platforms that they are deploying on (notably WinXP and
MacOSX, where Tk previously didn't look at all good.)

People using Motif desktops are much less affected by this, of course,
but they're becoming much rarer these days.

Now, given that this is not yet a TIP, but just an extension, that it is
answers a number of people's real needs, and that it is still under
active development, what is there to complain about? The "poll real
users" stage is *now*, though a dialog about how to achieve these things
is more useful than just a straight poll.

> The whole way this is
> being forced down our throats just because the Tcl developers think
> it is a more pure, abstract, model more aligned with the
> cutting-edge of academic software theory, without any regard for the
> real needs of real users strikes me as misguided, and points to a
> defect in the TIP process.

Woah! Where did that come from? OK, the Tile work is being done by a
number of Tclers who are also well engaged with the TIP process, but
that does not at all mean that it is yet part of that process.

> To me, it appears that the TIP committee
> seems to have lost sight of the fact that Tcl/Tk was not set up to
> be the next Algol or the next Ada, models of academic Computer
> Science purity, but to be a practical way to quickly solve
> real-world problems.

But the Real World point is that you do *not* need to use themed widgets
in your GUI. The existing Tk widgets will continue to be supported
precisely because there are a number of situations where the themed
widgets will perform poorly, and an explicit goal will be allowing the
standard and the themed widget sets to work together nicely if
developers so choose to do so. For some things, this will make a huge
amount of sense anyway.

It feels to me like you've not just read between the lines, but also
between the lines that you'd interpolated in. The reality is not what
you perceive it to be, and this works in your favour as a developer.
Furthermore, this themed-GUI work is being driven particularly by
commercial developers. (FWIW, as one of the more academic TCT members
and Tcl maintainers, I'm focussed more on doing cool things with Tcl's
internals. With Tile I really just cheer people on and make suggestions
for things they can do with the Tcl-level API to it.)

OTOH, adapting a complex GUI to use the themed widgets will be a
difficult task because theming is a different approach to GUI
construction. Ideally, it is about describing a GUI in higher-level
terms (e.g. it's not just a button with some funny options, but it's a
toolbar button, or it's a button with a flashing alert attached, or
something like that.) This process requires developer insight; it can't
be done automatically.

Your comments overall are IMO out of line. They appear to be based on
misunderstandings that have little to do with reality.

Donal (We're not taking the standard Tk widgets away, we're adding new
        things to them along-side. Got it?)