>audio does not suffice if I cannot hear it.
>
... and even if you can hear it, once you go beyond a very small handful of terms and variables it very rapidly degrades into a useless stream of dribble.
Certainly not suitable for anything but casual listening.  Audio plots of functions can, on the other hand, be extremely useful.

I believe that the plain ASCII text solution that Alan has decided to adopt is the most prudent at this time.  As Richard said, technology and ideas are not the
limiting factor here -- money is.  I have looked at a whole lot of web sites with mathematical content, and I can say that the Drexel Math Forum pages are the
only ones that I have been able to use without extraordinary means.  I don't have Duxbury, and I find LaTeX to be possibly the most inaccessible document
format I have ever dealt with (makes PDF seem like a walk in the park).  The converters like tex2html produce such horrible results that they are hardly
worth the effort, and this is the principal difference between the plain text solution and the more elaborate ones like MathML, LaTeX, MathWrite, etc. -- in
order to be able to use it, one only needs a screen reader or braille display with no need to run external applications, produce modified files, or use
specialized hardware.

I should clarify that I am talking about access as it pertains to blind users as this is currently the hardest nut to crack.  I imagine that a certain sector of the
learning disabled community could also benefit for some augmentation to standard graphical images, but as far as I know there is no application that will
verbally render mathematical content with synchronized visual highlighting while also giving the user control over the speed and granularity of what is to be
spoken.  Without this control, an audio rendering to a learning disabled person will quickly reach its limits of usability.

Electronic representation and manipulation of mathematical content seems to be one of those areas where the gap between "accessibility" and usability is
still quite far apart both in implementation, and in the general awareness of the scientific community.  John Gardner and a few others are making great
contributions and advances to this field, but we still have a way to go before there is anything that I would call truly intuitive, interactive, and accessible that
doesn't require extraordinary means to use.

-- Jim