[Tagging] Use of highway=track vs highway=service cemeteries, parks, allotment gardens, golf courses, and recreation areas

[Kevin Kenny]

On Fri, Feb 26, 2021 at 11:22 AM Christoph Hormann <osm at imagico.de> wrote:

> On Thursday 25 February 2021, Bert -Araali- Van Opstal wrote:
> > The standard highway tagging is mainly based on the physical
> > appearance.  Not the functional or just partly the socio-economic
> > importance.
>
> The opposite is the case - standard highway tagging - with the exception
> of highway=motorway - is almost purely functional in OSM.  There are
> some exceptions from that in local practice (like distinction between
> highway=trunk and highway=primary in Germany for example).  But overall
> all of the main road classes are overwhelmingly used with a functional
> semantic delineation.  This is also something data users (both
> cartographic and routing) massively rely on.


I think you're overstating your case just a little bit.

You often have to go to a global level - beyond the scale of a given
country or region -  before there's a functional difference between the two
interpretations. Most jurisdictions are sensible enough that the roads of
greater social or economic importance are also the roads that are built to
handle faster, higher-volume traffic. It would be uneconomical to do
otherwise. (Of course, politics being politics, there are exceptions to the
general rule...)  Where different jurisdictions differ, it is generally in
the scale of construction that they can afford. In one region, a primary
highway may be four lanes of asphalt as smooth as a baby's bottom, in
another, a primary road may be relatively dodgy compacted gravel and fine
material. Nevertheless, virtually everywhere, the more important roads are
better constructed than the less important roads in a given region.

Motor routers are generally written to prefer motorways to primary roads,
primary to secondary, and so on. Whether the hierarchy is based on social
and economic importance or on physical characteristics is therefore to have
relatively little impact on routing decisions simply because roads of more
economic importance are generally built to a higher standard. Switching the
interpretation between one and the other is unlikely to be the routing
catastrophe that you seem to imagine.

Where the choice of importance vs. construction does have an impact is on
the cartography, but that's an unsolved problem at the present time.  In a
large and extremely sparse area such as is found in much of North America,
the cartographic problem is that once the map is at a zoom level where the
minor roads become visible, it's at too large a scale to present
understandable context. Artificially elevating the importance of the roads
to make them display at a smaller scale isn't right, either. The minor
roads truly are of low economic importance, simply because there are no
high-value assets in the hinterland that they serve. They're no better
constructed than their importance warrants. By contrast, urban arterial
streets, which aroud here are often tagged 'tertiary', make the map
unreadably cluttered at the same zoom level. The 'tertiary' tag is
nevertheless warranted, because these arterials are justifiably viewed as
being of greater economic importance. Because the natural scale of city vs.
coutryside varies so much, neither hierarchical interpretation solves the
cartographic problem.

I've experimented off and on with a renderer that adjusts the threshold for
zoom level at which objects appear, based on intersecting the object's
geometry with the 'built-up areas' on Natural Earth. My current
implementation doesn't work too badly for filtering points of interest, but
it is nowhere near to being ready for wide use on linear features.  It
introduces too many anomalies near the boundary of the Natural Earth
polygons to be attractive or informative.

The question that motivated this over-long thread is, in any case, not
attacking the hierarchy of motorway, trunk, primary, secondary, tertiary,
unclassified, residential, service.  Really, the only places where there's
been any confusion - and there's repeated confusion - are that North
Americans don't have a single definition of 'trunk road', which is a legal
classification in the UK that doesn't really map to our highway network
unambiguously, and that we appear to have a greater and more nebulous
variety of relatively unimproved roads of little economic importance, for
which a firm classification of 'unclassified', 'residential', 'service' and
'track' is difficult or impossible to assign without research beyond what a
mapper in the field can achieve. I argue that fuzziness at the category
boundaries - particularly at the extreme high and extreme low ends of
highway quality or importance - is likely to  have only trivial impacts on
either rendering or routing.

I realize that you are considerably less willing to tolerate ambiguity than
I am - yours has been one of the voices that tends to answer the corner
cases with "if you can't determine a bright line for a classification,
don't map it."  Nevertheless, I think that the consequences of ignoring
this particular cry are minimal. My impression is that many others in North
America agree with me.

-- 
73 de ke9tv/2, Kevin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20210226/7b7512a4/attachment-0001.htm>