Development Strategy for Cross Platform Educational Software

What is the optimal development strategy for writing cross platform
educational software?  The tricky part is that it must run on both
PC's and Macs, and have an online version that doesn't require
administrative privileges on the user's machine.  I would also like to
leave open the option of porting to OpenGL/DirectX video game
architecture.

To make it slightly more complex, the art used in the software (and
still shots of the interactive animations) will be used as
illustrations in an accompanying textbook.

The app is strictly 2D.  The GUI will have multiple panes, interactive
animations and graphing tools.  As far as I can tell, it seems the
best approach would be to develop the application for three separate
platforms:  Macromedia Flash for the online version, Java SWT for the
desktop version, and use a third party game development engine for the
OpenGL/ActiveX video game version.

If this is true, what's the best way to organize development for those
3 target platforms?

It seems that the least expensive one to develop is Flash because many
Flash programmers are also artists, and they can develop an entire
application so rapidly.  But  what steps can be taken to make it
easier porting the video game from Flash to Java SWT and
OpenGL/ActiveX?  And how should the art be ported to a publishing
application?

Why can't you just use Flash for all of your versions?  If the
processing needs of the software truly require DirectX, then Flash and
Java will never suffice.  If Flash is sufficient, then why make a Java
or DirectX version?  DirectX development can be very time consuming
unless you have a game engine built on that technology that you are
already working from.  The DirectX "draw a box on the screen" sample
program is like a hundred lines of not-so-intuitive code.  I've seen
plenty of children's games and educational software written entirely
in Flash.

Maybe I'm missing something but it doesn't sound like you have any
needs that truly require that you do parallel development in three
completely different software architectures.  If you absolutely must
go with your original plan, I'd at least suggest starting with the
simplest of the three first so if your project runs out of steam you
still have something to show for your effort.

The program will include very complex features.  Flash performance
becomes slow, unresponsive, and chunky with only medium complexity. 
Java SWT allows for rapid development of much more complex GUI
features such as dockable windows, file trees, diagram widgets, and OS
integration.  The idea is to pack as much functionality into the Flash
version until it becomes unwieldy for both use and development.  Then
add extra functionality only to the Java SWT version.  The option of
using DirectX is left open so that video game style responsiveness is
available for a version just with features that would benefit from
intense graphics and rapid user interaction.

And I thought I was the only one to use the word "chunky" in
describing poorly performing systems!  If you need "dockable windows,
file trees, diagram widgets..." I can definitely see why you'd want to
use Java, since you'd be able to harness a componentized software
infrastructure.  Am still a little skeptical about DirectX, especially
as far as dockable windows, file trees and diagram widgets.  I'm sure
you can do it, but it'll take some real dedication.

Since you know that the DirectX step will be the hardest, I'd design
from the beginning to minimize the possibility of doing a total
rewrite during that port.  First figure out which game engine you want
to use.  Architect your game with the game engine but don't write it. 
Then use that software architecture to begin building in Flash, then
Java.  At least with this approach, when you get to the hard part,
you've made sure that you've got the wind at your back, so to speak.