[Accessibility] Mapping complex road junctions with pedestrian crossings

[John Sturdy]

Hi,

before I raise something for discussion I'd better introduce myself
--- I'm a sighted mapper (and professional software developer) and I
have a blind friend who's already on at least as many mailing lists as
she wants, so when I got into OSM I said I'd watch out for interesting
developments... and I have some experience of helping her learn new
routes, so I have a second-hand idea of what might be useful...

The question has come up, on my regional mailing list, of how much
detail should be included in mapping a complex road junction, one of
the examples given having several pedestrian crossings with "islands"
(I think some people call them "refuges", that might be a US term).  I
suggested that the islands should be mapped explicitly, rather than
being left as a gap between two pieces of road, and that the fences on
them should be mapped, with a specific view to this being useful to
the blind.

But on further thought, I've realized that it might be better to map
the pedestrian routes through such a crossing, so that it is explicit
that, for example, on crossing one carriageway and reaching an island,
that you have to turn left, walk to the other end of the island, press
another crossing button, and turn right to cross the second
carriageway.

I see that Lulu-Ann has already proposed a related tag,
land_use=traffic_island (presumably an area), and of course there's
already highway=crossing and crossing=island (which I think are
expected to be nodes); what I'm suggesting is a type of way, that
shows the exact route through a complex crossing (it is common in the
UK for the two crossings onto a mid-road island not to be facing each
other, I think this is typically if they have separately signalled
traffic lights, so you can't walk across in a straight line into a
stream of traffic controlled by a different light from the one that
has let you cross the first part).

We could just mark them as "highway=footway" but there's a risk that
non-blindness-aware mappers might remove them as clutter; I think we
should use a separate type, that perhaps visual renderers might omit
or make inconspicuous.

I suggest that tactile paving and crossing buttons could be used as
nodes on these ways, so that the way will hold together all the pieces
of information useful for a blind person using that crossing.

Anyway, I thought I'd better ask where I can get some answers from
people with first-hand experience of non-visual navigation, rather
than just guessing from second-hand experience.  (I'll ask my blind
friend too, she's not yet involved with OSM though.)

__John