On Wed, Feb 20, 2019 at 8:08 AM Florian Lohoff <f...@zz.de> wrote:
> Why not? This enables a routing engine to assume different
> characteristics of roads.

To make a routing engine such as OSRM and GraphHopper change its
default assumptions about a road type, please describe the road in
further detail using maxspeed=*, surface=*, smoothness=* and also the
access=* tag hierarchy.

Commercial maps usually do not change highway type on roads that go
through urban areas even if they change their physical characteristics
and legal requirements. Such a situation might be rare in Germany but
it happens in smaller cities in many developing countries (mostly
attributed to disorderly urban growth).

The purpose of highway classification is to clearly identify
thoroughfares (in German,  Durchgangsstraße), that is, the main routes
between places. The main article on the highway tag [1] starts from
this idea (highway=primary connects large towns, highway=secondary
connects small towns, etc.) and then asks mappers to choose a similar
local/recognizable system of roads that approximates this result (eg.
choosing ways of a particular official category, or with special
signage).

The road's classification is not simply a summary of the road's
physical characteristics, even though both are usually correlated and
several country mapper communities have adopted such qualities to
establish their highway classification rules in an attempt to improve
verifiability [2] and reduce edit wars. As an example that physical
qualities do not strictly determine highway class, in Canada trunks do
not have to be paved. [3]

Applications expect each level of the street mesh to be as complete
and connected as possible. Some routing engines (especially older
engines and software running on embedded devices with low CPU/memory,
not the case of OSRM and GraphHopper) employ heuristics that
prioritize following streets of higher classification first. In that
sense, highway=unclassified would work like a level between
residential and tertiary, that is, this kind of routing engine would
avoid computing routes through residential streets unless it cannot
find a route using higher road classes. So if an highway=unclassified
route is interrupted by a string of highway=residential in an urban
area, the engine will avoid considering the through route, it will do
so only after exploring all other alternatives of higher class, and
that may result in a much larger route around that area in the highway
mesh.

I don't know any rendering style that currently differentiates between
residential and unclassified. In functional classification both are
types of local roads. [4]

> > And usually a connecting road from outside a city limit has at least a bit
> > more traffic as an inner-city-only residential.
>
> Have you had a look at the original example images for an unclassified?
>
> https://wiki.openstreetmap.org/wiki/File:Highway_unclassified-photo.jpg
>
> I would not expect more traffic here - I'd expect less.

You can't judge traffic from a single still picture. While I don't
know Germany in such detail, this type of road sometimes will have a
different type of traffic, like fewer cars and more trucks or
tractors, depending on where it appears.

> > So the conclusion an unclassified has a bit higher priority than a
> > residential is not far from reality.
>
> Not in my reality and not in the original OSMs reality. Yes - through
> misleading statements in the German article this might have influenced
> at least the German community to assume otherwise - This is why
> i request clarification.

I'm Brazilian (and mapping here) and I actually find the German
classification rules quite reasonable and inspiring, including this
rule about unclassified=*. As far as I know, as you have said, the
German rules were based on English rules, for which
highway=unclassified simply meant "public roads with no official
class" [5]. It is not very intuitive for those classifying roads
without considering their context though. Rules based solely on
physical appearance often lead to fragmentary highway meshes. See [6]
for a very long discussion on the matter. Not everybody agrees that
fragmentation is a problem.

> > Otherwise there is often the problem to tag the main access roads inside a
> > bigger residential area.
>
> The region where i map mostly we agreed that we may tag roads with clear
> interconnecting character and wider lanes with one class higher than
> they would have by assuming the strict classification. We agreed
> that the causes by which we tag higher be placed in a note= tag on
> the road.

Any information you provide on why you chose a particular
classification helps others verifying your changes and avoids edit
wars. I often use source:highway=* for that purpose, as source:* is
the default tag namespace to express where any tag value came from.
[7]

> > The practice to tag those as unclassified for a bit higher priority may not
> > be optimal - but suitable.
> >
> > This discussion - and usage - is some years old now - and I thought you had
> > at least knowledge of it from the german forum.
>
> My knowledge and usage predates the German Forum by years - I was astonished
> finding statements in the German article for unclassified which do not match
> (but oppose) the English versions which i typically use and prefer.

Everyone's opinion counts, but the German article has had this
information since May 2009. It would be best to check if what's on the
map today is closer to what's in the German wiki or to what you have
been doing.

Regards,

[1] https://wiki.openstreetmap.org/wiki/Key:highway
[2] https://wiki.openstreetmap.org/wiki/Verifiability
[3] https://wiki.openstreetmap.org/wiki/Canadian_tagging_guidelines#Trunk
[4] https://en.wikipedia.org/wiki/Functional_classification
[5] https://en.wikipedia.org/wiki/Hierarchy_of_roads#Unclassified
[6] https://lists.openstreetmap.org/pipermail/talk/2018-February/080155.html
[7] https://wiki.openstreetmap.org/wiki/Key:source

-- 
Fernando Trebien

_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging

Reply via email to