[Tagging] sidewalks

Peter Wendorff wendorff at uni-paderborn.de
Thu Aug 26 22:21:26 BST 2010

  Hi David.
Thanks for your reply - I'll comment your (shortened) mail in between.

On 26.08.2010 22:47, David ``Smith'' wrote:
> On Thu, Aug 26, 2010 at 4:25 PM, Peter Wendorff
> <wendorff at uni-paderborn.de>  wrote:
> Makes sense to me.  You might consider one minor modification, though:
> maybe we could add a couple of disambiguation words to the name for
> sidewalks.  For example, an east-west street called Drury Lane might
> have sidewalks called Drury Lane North Sidewalk and Drury Lane South
> Sidewalk.  This disambiguates the sidewalks from the street and from
> each other.
Well... if a sidewalk has a dedicated name I would have mapped it as 
separate way without tagging it as a sidewalk, as it's another way by name.
Adding "south" "north" etc. to sidewalks is not a good idea. If an 
application needs this data, it has to get the geometric relation 
between sidewalk and street to add north/south/west/east or - if 
preferred left/right (depending on the walking direction).
> Some other points that may well have already crossed your mind:
> * If a street has its sidewalks mapped separately, the street itself
> should probably be tagged with foot=no.
That's possible and should be recommended, yes.
> * This provides an opportunity to characterize each sidewalk/street
> crossing (crosswalk).  On the node where they intersect, tags could be
> used to indicate whether there is a dedicated crosswalk signal tied to
> the traffic signal, or if the traffic signal implicitly governs the
> crosswalk too, or if there is no such signal.  There can be tagging
> for audible signals.  Of course, if there simply is no crosswalk on
> that side of the intersection, then the sidewalk shouldn't be mapped
> between the two corners concerned.  This information can be extremely
> useful to navigation for the blind.
;) I know.
I tag crossings as a line of 5(!) nodes - let's call them A,B,C,B,A.
C is the node shared between street and crossing footway. it's tagged as
crossing_ref=zebra|pelican... if one fits

if crossing=traffic_signals is applied, additional tags are:

The nodes B describe the border between sidewalk and street. Thus they 
are tagged as
where * can be an arbitrary height in metric dimension. The numbering 
scheme is proposed by the wheelchair-routing people, because wheelchair 
users need to know the exact height of the curb to decide wether they 
can drive up/down there. For blind people the reduced tagging scheme is 
enough and that's easy for the mappers as it's possible to collect these 
data without measuring (even from the bus).
primitive is applied where there is a tactile paving, but not made of 
"THE" tactile paving stones. Often small cobblestones are used instead.

The A nodes are the connectors to the sidewalks on both sides of the 
It's not correct to merge A and B node to one node, because that could 
lead to a tactile paving not connected to the specific crossing.

I hope, I didn't forget parts of the crossing tagging I apply currently ;)
> * Instructions generated by a blind-navigation program, particularly
> when based on data with streets and sidewalks separated, should almost
> certainly include instructions like "Cross Drury Lane".
Yes, exactly - and therefore explicit tagged crossings are necessary. 
Crossing a street is really dangerous in most cases for blind people.
Being able to get best crossing places at the street is the one and only 
solution to get best possible security at that point.

If you want, look at the streets and crossings I mapped here. Most 
footways should be tagged no (the sidewalk scheme is not applied yet, 
sidewalks are unnamed there). Link: 
Note: the footways often are not visible because they are hidden by the 
streets. I tried to locate footways in a distance to the streets as they 
are (street.width/2+sidewalk.width/2), but streets are rendered wider 
than they are in most cases.


More information about the Tagging mailing list