[OSM-talk] Sidewalks

Frederik Ramm frederik at remote.org
Sat Apr 25 12:16:57 UTC 2015


On 04/25/2015 12:33 PM, Serge Wroclawski wrote:
>> To improve that I would like to encourage mappers to give separately mapped
>> footways their proper name instead of leaving them without name.

> Why do that instead of just adding a single tag to the road?

Roland's use case is routing for pedestrians.

If a road has a tag indicating "this road has a sidewalk" but the
sidewalk is not mapped separately, then the router will lead the
pedestrian onto the road which is ok.

If however the sidewalk is - for whatever reason - mapped as a separate
highway=footway, then today it will often be un-named, which leads to
the routing engine generating instructions like "follow un-named footway
for 2 miles" when instead it should be "follow (footway along) Main
Street for 2 miles".

Roland's point is that it is too complicated for a routing engine to
guess that one un-named footway is really part of "Main Street" and
should be announced as such, whereas another un-named footway might
really be nameless.

His initial suggestion was to simply add the street name to every
separately mapped sidewalk. This was criticised because it would likely
lead to labeling chaos on the rendering side (with renderers then having
to drop footway labeling altogether or implement complex rules like
"don't label this if there's a roughly parallel street of the same name"
or so). Roland then amended his suggestion to say that if a sidewalk
receives (a copy of) the name of the street then it should also be
tagged footway=sidewalk so that renderers could choose to omit only the
names of these (and not all footways).

Personally I am still doubtful whether the sidewalk next to "X Street"
really has the name "X Street" but at least the addition of
footway=sidewalk would let users decide how to handle it. For example, a
geocoder would likely want to omit indexing footway=sidewalk for forward


Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"

More information about the talk mailing list