Another caveat is that I haven't had much experience with GIS software (with the exception of MapPoint back in the early 00s). I'm sorry if this all comes across as rather rudimentary; I'm only trying to help.

At the risk of becoming 'Captain Obvious", I would suggest placing some sort of boundary around what are you are going to work with in order to make something accessible. In other words, limit your map to a certain area and build only upon that area.

From there, I would think about making lists for the streets, using the main streets as the root of those lists. This would allow you to tag coordinates in the [label] alt text a little easier to make them correspond with searches. Unfortunately, only after one massive list (and subsequent sub-lists) is generated, can we really begin to address the bulk of the issues of the Checkpoints.

Structuring as lists allows for a bit more leeway in terms of directional components to screen readers. We can really make use of [label], [lbody], and [caption] references as well as alt tags for [LI] and [UL]. Having a visual aid (e.g. a Legend) for myself helped to differentiate the various list labels, for example, which I used to signify one type of subway stop or the other. Similarly, you could use list labels to point out block information and Caption tags to explain coordinates or intersections.

I think this would solve some of Checkpoint 1, 2 & 8. I think identifying the function of the map would probably dictate how you solve Checkpoint 9 & 12. If an enduser is using the map to get from point A to point B, I assume that the search term she or he is using to find something will check the database for an acceptable match, and that query would return a value that is equal to coordinates that match up with those in the list mentioned above. Therefore, the information that would be returned would be generated from the list and you wouldn't need a mouse to use that. I suspect that some script would need to translate coordinates to a Left or Right turn and how many blocks.

Which leads me to Checkpoint 6. I don't think there is a way to avoid using scripts, because the nature of all this requires the user to ask the server something and have some information retrieved to explain it back to the user.

I think if you are able to make a more accessible map, it would be better than making a fully 508 or WCAG compliant map, understanding that maps are still considered too much of a burden. However, you'd probably be the first to figure this out, and you'd be able to attract that many more accessible users.

I hope this was helpful.

Jonathan Metz http://www.jonathanmetz.com

very interesting article