Re: [OSM-talk-be] Fwd: [OSM-talk] Highway=trunk : harmonization between countries ?

[Tim Couwelier]

There's always an inherit 'gap' between 'what does government intend the
road for' and 'how does the road actually look'.
Terms such as 'primary' and 'secondary' roads have meaning in planning
context.
In Flanders we have the 'Ruimtelijk Structuurplan Vlaanderen' which
classifies which roads are 'hoofdwegen' and which are primary (divided in
classes I and II)
Each Province has a 'Provinciaal Ruimtelijk Structuurplan', which
clasfficies which roads are secondary. (divided in classes I, II and III)
Each city/town is also supposed to have a 'Gemeentelijk Mobiliteitsplan'
which states which classification 'local' roads get, divided in classes I,
II and II)

There's potential for a very close match:

E and A roads are 'hoofdwegen'  => Tag as 'motorway'
The primary roads (regardless of their 'collecting' or 'connecting'
subtyping)  => tag as primary
The secondary roads (also regardless of their subtyping) => tag as secondary
The local roads types I and II => tag as tertiary
The local roads types III (aka 'the rest') => multitude of tagging options.

With the current tagging system and how it's applied in Belgium, we go
somewhere in between, but it's fairly 'clean' as it stands.
We don't look at what the government says, we go by 'how it looks' and link
that to the road numbering system for the' gewestwegen'.
Key point to decide on is if we SHOULD bother with 'intent' rather then
'reality', as 'mapping what's on the ground' is a basic principle.


Relating back to the post Joost distributed:
I do agree with most of the points, although 'trunk' is the odd part out.
'trunk' we don't use as a hierarchical classification, but to point out
it's a strech with a certain setup, i.e. forbidden for cyclists and such.


To end I'll repeat the example I've given in the riot channel about the
subtle difference between 'intended' and 'assumed':
The R32 ringroad around Roeselare.
Given it's a 'ringroad' it's classified in OSM as 'primary' all around.
But from the 'planning context' viewpoint, only the last stretches towards
the E403 are 'primary', and the majority is only 'secondary'. While the
road goes around Roeselare, it's function is to get people from and towards
the E403 'in either direction'. Due to the E403 being present, it's never
the intention to use the R32 to go around the entire end, as the E403 helps
cover that function.

PS:
If you have a look at said area, you'll also notice a part of 'trunk'.
Rendering wise, it 'feels' like the classification is different, and in
reality it looks different, but its function isn't really different at all.
Along with the aforementioned nuance primary/secondary, it's a second
example on how you could interpret on the same road.


2018-02-22 21:45 GMT+01:00 joost schouppe :

> Hi,
>
> Not wanting to change current consensus in Belgium, but I wonder how close
> this would be to current mapping practice in Belgium, and if it would be a
> way of thinking that could help in some current edge cases.
>
> -- Forwarded message --
> From: Fernando Trebien 
> Date: 2018-02-15 19:14 GMT+01:00
> Subject: Re: [OSM-talk] Highway=trunk : harmonization between countries ?
> To: t...@openstreetmap.org
>
>
> Landing on this discussion several months late. I've just heard of it
> by reading a wiki talk page [1].
>
> Since 13 February 2009, the wiki [2] criticises highway classification
> as problematic/unverifiable. This has also been subject to a lot of
> controversy (and edit wars) in my local community (Brazil), especially
> regarding the effect of (lack of) pavement.
>
> In trying to achieve greater consensus some years ago, I decided to
> seek opinions elsewhere and finally I arrived at this scheme [3] which
> I think is very useful, if not perfect yet. It can be easily
> summarised like this:
> - trunk: best routes between large/important cities
> - primary: best routes between cities and above
> - secondary: best routes between towns/suburbs and above
> - tertiary: best routes between villages/neighbourhoods and above
> - unclassified: best routes between other place=* and above
>
> For example, the best route between two villages would be at least
> tertiary. So would be the best route between a village and a town or a
> city. Parts of this route might have a higher class in case they are
> part of a route between more important places.
>
> It surely raises the problem of determining optimal routes. Maybe a
> sensible criterion would be average travel time without traffic
> congestion. A number of vehicles may be selected for this average -
> could be motorcycle+car+bus+truck, or simply car+truck.
>
> Early results in my area [4, in Portuguese] seem promising and have
> produced more consensus than any previous proposals. To me, this
> method seems to:
> - resist alternations in classification along the same road
> - work across borders (where classification discontinuities are
> expected because each country is 

[OSM-talk-be] Fwd: [OSM-talk] Highway=trunk : harmonization between countries ?

[joost schouppe]

Hi,

Not wanting to change current consensus in Belgium, but I wonder how close
this would be to current mapping practice in Belgium, and if it would be a
way of thinking that could help in some current edge cases.

-- Forwarded message --
From: Fernando Trebien 
Date: 2018-02-15 19:14 GMT+01:00
Subject: Re: [OSM-talk] Highway=trunk : harmonization between countries ?
To: t...@openstreetmap.org


Landing on this discussion several months late. I've just heard of it
by reading a wiki talk page [1].

Since 13 February 2009, the wiki [2] criticises highway classification
as problematic/unverifiable. This has also been subject to a lot of
controversy (and edit wars) in my local community (Brazil), especially
regarding the effect of (lack of) pavement.

In trying to achieve greater consensus some years ago, I decided to
seek opinions elsewhere and finally I arrived at this scheme [3] which
I think is very useful, if not perfect yet. It can be easily
summarised like this:
- trunk: best routes between large/important cities
- primary: best routes between cities and above
- secondary: best routes between towns/suburbs and above
- tertiary: best routes between villages/neighbourhoods and above
- unclassified: best routes between other place=* and above

For example, the best route between two villages would be at least
tertiary. So would be the best route between a village and a town or a
city. Parts of this route might have a higher class in case they are
part of a route between more important places.

It surely raises the problem of determining optimal routes. Maybe a
sensible criterion would be average travel time without traffic
congestion. A number of vehicles may be selected for this average -
could be motorcycle+car+bus+truck, or simply car+truck.

Early results in my area [4, in Portuguese] seem promising and have
produced more consensus than any previous proposals. To me, this
method seems to:
- resist alternations in classification along the same road
- work across borders (where classification discontinuities are
expected because each country is using different classification
criteria)
- account for road network topology
- work in countries with mostly precarious/unpaved roads or
without/unknown official highway classes
- work between settlements as well as within settlements

Borderline cases are probably inescapable in any system that does not
use solely criteria that are directly verifiable - from the ground, or
from the law. Maybe, in certain developed countries, the system is so
well organized that merely checking signs/laws is sufficient. That
does not mean it is like that everywhere on the planet.

OSM has so far received a lot of input from communities in developed
countries (mostly Europe, North America and Australia) and hasn't
given much attention to less developed/organized countries. What comes
closest to this is what the HOT Team does, but the judgment of road
classification one can do from satellite images in a foreign country
is much more limited than the criteria that have been raised in this
thread so far.

I wouldn't endorse tags such as maxspeed:practical due to lack of
verifiability (it should be obvious that different types of vehicles
would achieve different practical speeds). It is better to use the
legal speed in maxspeed=* and describe the practical reason for a
lower speed using surface=*, smoothness=*, and, who knows, maybe the
not yet approved hazard=* [5] (though that is intended for signed
hazards, not subjective/opinionated hazards).

For the sake of long-term sanity, I also wouldn't mix the purpose of
one tag with the purpose of other tags. To describe the surface, there
is surface=*, smoothness=* and tracktype=*. To describe access rights,
there is access=*, foot=*, bicycle=*, motor_vehicle=*, etc. To
describe legal speed, maxspeed=*. To describe curves, there's
geometry.

Purpose, perhaps, is the main issue. What is the purpose of highway
classification? Is it to save us the work of adding extra tags? Is it
to allow the renderer to produce a cleaner output at low zoom levels?
Is it to allow routers to assume default speeds? Maybe to guide their
routing heuristics? Is it to express some sort of importance? If so,
by which perspective - urbanistic, traffic engineering, movement,
commercial value, cultural/fame, historic, some combination of those?
Should the purpose be the same in every country?

It may be interesting to also discuss the classification adopted by
other maps. I don't have a reference for Google (originally TeleAtlas)
or Here.com (originally Navteq), but Waze publishes its per-country
road classification criteria in its wiki. [6-16]

[1] https://wiki.openstreetmap.org/wiki/Talk:Tag:highway%
3Dtrunk#change_.22high_performance.22_to_.22high_importance.22
[2] https://wiki.openstreetmap.org/wiki/Verifiability#Problematic_tags
[3] https://wiki.openstreetmap.org/wiki/User:Ftrebien/Drafts/