What advice would you give to someone developing a windowing system that allows the arbitrary placement of "layers" in a 3D space for sighted users, where each layer represents content such as an HTML document? I know where I can find basic information for incorporating existing affordances like those in WAI-ARIA, but I haven't a clue where one would start learning the basic underlying principles for designing any fundamentally graphical system for eventual use by the blind.

FWIW, I'm one of the software engineers at famo.us and I'm genuinely curious about knowing how to start designing accessibility features normally found in operating systems, rich application frameworks and windowing systems (qt, kde, gnome, cocoa, wpf, UIKit, AppKit, etc.) from first principles.

At the end of the day, at famo.us we're fundamentally exploring and implementing concepts in client-side JavaScript that have historically been the domain of the user agent makers and been implemented directly in browsers and accessibility tools like screen readers. With that in mind, do you know of people/projects/efforts to implement accessibility features directly in JavaScript? If not, what kind of accessibility features would you like to see that could be implemented in client-side JavaScript?

At the end of the day, HTML Documents have always been linear in nature, but applications are fundamentally non-linear, so a lot of the concepts I'm familiar with don't exactly map to what we're doing with scene graphs. Furthermore, the trend for interface designers to explore z-depth as an affordance makes things simpler for sighted users, but I would imagine complicates things for visually impaired users.

Lastly, besides the recently founded "a11ybay", are there any other notable developer communities/meetups you know of in or around San Francisco where one can meet engineers specializing in accessibility.

related:

[0] The Extensible Web Manifesto http://extensiblewebmanifesto.org/

[1] a11ybay http://www.meetup.com/a11ybay/