[Tagging] Pedestrian access tagging

Brian M. Sperlongano zelonewolf at gmail.com
Fri Jan 22 21:55:33 UTC 2021

I operate a site which tracks completeness for runners/joggers attempting
to run every every street in their city.  As a data consumer, I extract
streets in a city which are accessible to runners.  For example, I assume
that streets tagged access=private, foot=no, or highway=motorway are not
accessible to runners (among other tagging filters).

I often field questions from users when they encounter data which is
unexpected, such as "this is a private road" or "this road doesn't exist in
reality", and direct them to OSM to make the appropriate edits.

One category of road that I DO include is highway=trunk.  In many places,
it is just fine to run along the shoulders of trunk roads, and so I include
them by default (provided that they don't have other tagging such as
access=private or foot=no that would cause me to exclude them.

One of my users sent over this link which shows a road tagged highway=trunk
which is clearly not accessible to runners (or at least, those that are not


In my area, I would simply tag a road like this with foot=no, as any
pedestrian router would obey this tag, and no pedestrian router should
route over a road like this.

>From talking to other mappers, there may be a European/American divide over
how foot=no is applied.  The feedback I've heard rather consistently from
US-based mappers is that they apply a "reasonable person" standard to roads
like this, and tag foot=no any place where a reasonable person would
conclude that pedestrian access is physically impossible (as in this
example), in addition to places where it is signed as illegal.

The European view seems to be that foot=no is ONLY applied in places where
pedestrian access is signed as illegal or otherwise codified in law.  This
viewpoint goes on to say that routers should come up with some algorithm
that combines shoulder width, the presence of sidewalks, perhaps speed
limit, etc., to make a determination as to whether a road is traversable by

This strict viewpoint creates a considerable for any router trying to
determine the pedestrian suitability of a road.  The shoulder:width key is
barely used, and even the shoulder=* key overall has only 50K usages -
inconsequential when compared to the 1.3M usages of highway=trunk.  There's
no obvious way to tell whether a carriageway is boxed in with guardrails
without doing some kind of spatial query.  I could look at shoulder:width=0
+ sidewalk=no for this particular case, but that would inappropriately
exclude roads which have no paved shoulder but perhaps are low traffic or
for which you could step aside when traffic comes the other way -- a common
situation in more rural areas.

The documentation on this tag on the wiki takes the European view of
"legality only".  If my characterization of how foot=no is used in the US
is accurate, this wiki description is not appropriately descriptive of how
the tag is used globally.  Since foot=* is a de facto tag rather than an
approved proposal, I believe at a minimum a description of actual use is
appropriate there if there are true regional differences.

There is also a tag motorroad=yes which might work here, but it is unclear
as to whether this is a "strict legality" tag or represents a different

Interested in hearing community thoughts as to how you make use of the
foot=* tag in your area and recommendations on how to tackle this problem
generally from a data consumer perspective.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20210122/498a1ec8/attachment.htm>

More information about the Tagging mailing list