[Blindmath] How useful is a GUI to blind users?

[Jamal Mazrui]

Some miscellaneous points are as follows.

There were text mode, screen oriented user interfaces before graphical 
user interfaces, e.g., WordPerfect 5.1 was a highly popular DOS-based 
word processor among blind people.  In fact, I think we were, overall, 
more productive with it than any GUI word processor, like MS Word,  or 
command-oriented word processor, like VIM.

Those text mode, screen oriented interfaces could be highly accessible 
with a screen reader because it knew precisely what character (and 
color) was located in every position of a grid with 25 rows and 80 
columns.  The point of focus was also precisely known -- what later 
became the caret in a GUI.  Accessibility of GUIs has been a problem in 
large part because a screen reader has had to build an "off screen 
model" of what text is on the screen, since by the time text is 
displayed on the screen, it is pixels not characters.  A GUI screen 
reader has traditionally had to intercept every operating system call 
that writes to the screen, remember where the writing occurred, and 
adjust the model every time some other OS call implies a change in the 
visual presentation.

This process has been error prone, thus necessitating the need for an 
accessibility API by which the screen reader can interrogate the screen 
at any time.  Unfortunately, comprehensively implementing that 
accessibility API, on both the application and screen reader side, has 
also been error prone.  There are other reasons as well why GUIs have 
been difficult, including the emphasis on a mouse for input rather than 
the keyboard -- the former usually requiring fine eye-hand coordination 
to be practical.

Blind people, like sighted people, have a range of cognitive abilities.  
Some find the memorization of commands and syntax, which is generally 
required by a command-line interface, to be challenging and frustrating, 
as compared to a GUI and screen reader combination implemented 
accessibly, which support much more choice selection and safe 
exploration of options before executing a task.  Often, the most 
productive application will be one that offers both an accessible GUI 
for typical use and a well-designed command line interface for advanced use.

I am puzzled why Swing seems to be so integrated into popular Java 
libraries.  Do they not separate visual display from underlying 
functionality?  I thought there was generally a well-established 
principle in software development that it is best to implement a set of 
functionality in a way that can be connected to various presentations of 
user interface -- unless, of course, the library is specifically 
targeted to building user interfaces.

Regarding SWT, it is cross platform, not Windows only, though different 
binary files have to be distributed with other platforms.  I am a fan of 
wxWidgets, but agree that SWT is clearly more established in the Java 
community, and do not see any advantage that wx-based GUIs would offer 
over SWT.  Both GUI libraries are essentially wrappers over native 
controls of the platform.  Such controls tend to be the most accessible 
because they implement the primary accessibility API of the platform, 
being build by the developer of both the platform and accessibility API.

Jamal