Thesis Presentation to Sys Dev
				 5/2/88

1- 3 Introductions

   a- Me- User Consultant for 3 and a half years now.
	  Not a programmer by trade.  E.E. major, actually.
	  
   b- Andrew Toolkit developed at CMU after CMU decided to adopt X as
	their window system, instead of using their own.   Based mainly
	on their original work, be1.  It is a hierarchical, object-based
	high-level, event-driven toolkit.
	
   c- Olc code- various authors here at Athena:
	Win, Ken, Steve, Bill Saphir, Dan Morgan, et. al.

2- Ask how many people have used - real olc or olcr
				 - my olc or olcr

3- My thesis

   a- Learn about and evaluate the Andrew Toolkit through the porting of
	a non-window-based application to an Andrew Toolkit window-based
	application.  More specifically, I ported olc and olcr (the OLC
	client) programs to use the Andrew Toolit

   b- Through this porting to a window-based client, learn about what
	goes into the creation of a user interface

4- What I did...

   a- Created two client programs, olc and olcr.  First three pages in
	handout after cover page are screen dumps of the running
	clients.

   b- Object breakdown of olcrWindow (source code follows screen dumps)
	- editWindow
	  - titleBar
	  - lpair
	  - frame
	  - textview (scrollbar, 'cause application mode)
          - text
	  - buttons (source code follows olcrWindow source)
	- panelWindow
          - same as above, except +panel

      Total of 6 self created objects + lots of builtins (font, style,
		context, etc..);

   c- The clients actually work, an tie directly into olc-daemon on
	matisse.


5- What I learned...

   a- Andrew
	1- interaction manager queue not in sync with X queue;  have to
		be careful about things like the startup screen, etc..
	2- Toolkit is not strict about use of object data;  all object
		data can be accessed without object methods or
		macromethods; makes for sloppy programming
	3- Objects available directly from the toolkit are rather
		extensive, including a full-function multi-object editor
		object.
	4- To but down on overhead, sometimes better to modify their
		source code to create slightly different object, rather
		than subclass and override.
	5- lots of overhead; compiled and stripped client ~470K;
		xmh ~= 225K; unable to use saber and do development
		on RC; stuck with dbx (not that bad)
	6- interaction with X kinda slow (although I here of re-write);
		a few bugs in exposure-type events;  menus attrocious;
		turned off for applications
	7- VERY easy to create new objects once you get the hang of it
	8- Documentation not specific enough; need to look at sources
		and/or header files quite often when dealing with a new
		object of theirs
		i-  data types (e.h. rectangle)
		ii- explanation of routines

   b- User Interface
	1- Main Point- Can't really design on paper.  Need to look
		at on screen; need interation of design
	2- Difficult to control user interface in event driven system;
		Should I turn off all buttons when one is pressed?
		How to handle queries for info that user must type?
		How to display more than one message at a time.
	3- Not worth the time to turn off useless buttons (lot of
		time to figure out what is useless).  Better to
		let the user select, and tell him that it's not
		available because of current state.
	3- Different people have different ideas;  most consultants
		won't use my olcr because of speed; faster to type than
		use mouse to press buttons or move scroll bars;
		these clients probably a bigger win on the olc side
		of things (user)
	4- Button (and panel) interface better than menus; visually more
		real-estate dependent.  No hidden suprises.

6- Conclusion

	The Andrew Toolkit provided an very adequate means of porting
the olc and olcr clients to an easy-to-use window-based interface.  At
times, was frustrating because of lack of complete documentation.
Toolkit is rather complete, though.  User interface design is a tricky
matter.  Tied into the tools that you are using.  Need to make
sacrifices at the expense of speed.


7- Try it out.  Instructions on last page.

8- Questions?