Re: Setting window icons

[Havoc Pennington <hp redhat com>]

Peter Korn <korn sun com> writes: 
> This is tough.  One big question still to be figured-out is how to
> best address magnification.  One approach is to ask apps to render
> themselves larger (e.g.  130%, 160%, 200%), which is ideal for a lot
> of folks who consider themselves not disabled, just "getting on in
> life" and needing things a big bigger.  This includes virtually all
> of the "baby boom" generation here in the U.S. who are finding
> themselves considering glasses for the first time.

I find that people often switch X resolutions to 640x480 or so for
this - I've had several people at Linux shows come up and ask about
resolution switching giving this as the reason they wanted to do it.
Unfortunately you have to restart X to do this right now, unless you
use the XFree "virtual screen" feature that people hate. I know the X
team is working on an on-the-fly resize/rotate feature.

> For people with more serious visual impairments, rendering things
> 200% larger is nowhere near sufficient.  Commercial software
> magnification products typically will magnify from 2x to 16x, so one
> pixel becomes 4 or 9 or 16 or...  (and many of these products use
> "font" smoothing algorythms so the pixels aren't so blocky) - see
> Windows 98/2000 "Magnifier", or Macintosh CloseView for some
> examples.  For these users, they won't be that interested in
> applications re-rendering themselves - instead they will rely on
> their deskto/system-wide magnification product.

Right.

> So I think we should determine a small range of larger sizes, and
> also consider allowing the option of vector-based icons that render
> themselves as appropriate to the rectangle they are given.  Then a
> nicely behaved program could display larger icons as desired, either
> doing a best-fit with what's there, or generating from vectors.

To do this, or to have apps self-magnify as mentioned above, we need
vector-based drawing primitives in GDK as with Java. The whole toolkit
is pixel-based at the moment.

This is one of 3 or 4 jumbo-size long-term TODO items we've considered
making the focus of a future release (Pango was the big one for this
release), and is obviously a whole lot of work, both to implement the
rendering and to make the toolkit take advantage of it.

Havoc