[OSM-talk] Advanced highway tagging

[Andrew MacKinnon]

On Nov 20, 2007 9:20 AM, Sven Grüner <sven at schunterscouts.de> wrote:
> Andrew MacKinnon schrieb:
> > [...]
>
> This topic fits perfectly because we're heavily discussing the same on
> the German list for over a week by now. But of course this is a global
> issue because everywhere on the world the administrative classification
> doesn't reflect the actual importance/physical condition.
>
> So we have to define which of these is described by the highway-tag.
>
> Personnaly I prefer the administrative classification because IMHO it's
> the only way that two different mappers standing in front of the same
> road will always label it the same.

In my part of the world (Ontario, Canada), the administrative
classification is pretty useless for navigation. Here, there used to
be a fairly thorough network of (non-motorway) provincial highways but
in 1998 most of those were "downloaded", i.e. turned into local roads.
This doesn't mean that these roads are any less important than they
used to be, or that they are necessarily more important than roads
which were never provincial highways. Even some motorways are not
administered by the province, but rather by local authorities. In some
parts of the province, certain local roads are designated as numbered
"county roads", but this is far from consistent. If I were to map my
hometown (Kingston, Ontario) using only those administrative
classifications, I would end up with one motorway (blue), two
provincial highways starting at the edge of the city (red), a couple
of "county roads" (orange) and a sea of tertiary roads (coloured in a
very pale yellow) and residential roads. The administrative
classification should still be mapped, but I feel that it shouldn't be
used to colour the road on the main map (in most cases, anyway).
Google Maps, for example, generally ignores it.

For the importance aspect there are
> a lot of tags already which routers/renderers can take into account on
> their own preference: lanes=, maxspeed=,
> highway=stop/crossing/traffic_signals, width/est_width=, grade1-5, and
> more to be proposed. Addionionally there's a lot of info that can be
> extracted from the raw data like if it's a dual carriageway and the
> average distance between two junctions.
> Each user/vehicle might judge the quality/importance of a road
> differently and the mapper should not take this choice from him by
> assigning a strict importance-hierarchy.
My problem with using strictly objective measures of a road's
importance is that using only these tags to colour a road on a map
will result in a difficult-to-read map. If a single road changes from
a speed limit of 80km/h to a speed limit of 50km/h, etc. then back to
a speed limit of 80 km/h, most commercial maps would still show that
road as one colour to preserve readability, but OpenStreetMap would
probably switch from red to orange and back to red, making it
difficult to follow - renderers are stupid. My proposed "subjective"
level would be loosely based on the objective measures proposed above,
but it would ensure that roads aren't changing colour every 200m. The
more detailed tags should still be used for routing, but I feel that
it is problematic to use them for rendering purposes. It would be
possible to have a separate "subjective" importance level for each
class of road user.

As for the "zoom level" tag which I proposed, and which has already
been criticized, it isn't really a strict tag telling the renderer to
"render road X starting at zoom level Y" but rather a tag indicating
the "relative importance" of the road, i.e. how important a road is
relative to others in the area, so that minor roads in isolated areas
are shown at low zoom levels but not all the arterial roads in Central
London. I doubt it that this could be done automatically, and still
yield a usable map, so I suggest adding a hint to the renderer
(remember: renderers are stupid) if a road should show up at a higher
or lower zoom level than it does by default.