Philosophically, making mainstream devices accessible through a separate user agent is, in my opinion, far preferred to "Blind Guy Ghetto" products. With something like UPNP and V2 in the future, one can create an infinite series of user agents to fit every imaginable need without turning the original device into a Rube Goldberg machine. Imagine a thermostat that is perfectly usable by a blind person, a deaf person, Stephen Hawking, a textually impaired person a low vision person, someone with dexterity problems, etc. You would end up with a monster. If, on the other hand, a single thermostat was to have some form of wireless communication that exposed a UPNP protocol, a blind person could access it from a PAC Mate, laptop or something else on their home network. The low vision guy could run magnification on a tablet PC and achieve the same result. The thermostat manufacturer can add a $6 part and a little software (write it in India for cheap) and sell a million units with near universal accessibility to the product. Custom applications for blind people do have their place. No screen reader has ever come close to delivering a graphing calculator as good as Gardner's and no one has ever been quite as able to deliver arithmetic to blind kids as well as Ted Henter's new project. For the most part, though, the economies of scale tell me that the iPaq that sells millions of units per year is going to be profoundly less expensive than a PAC Mate which sells a thousand a month or so. I'm doing a lot of research on environmental accessibility (it's one of the primary reasons I left FS) and hope to see much of this showing up in hotels and assistive living centers in the next couple of years. I believe that all of the new work we'll be doing on this stuff will be covered by GPL but will be developed under GNU/Linux as well as a few flavors of Windows.
