[OSM-talk] definition of the main highway-tag

[Dermot McNally]

2009/7/31 Greg Troxel <gdt at ir.bbn.com>:

> I see Trunk as almost motorway, but a little deficient.  Definitely has
> to be divided by at least some concrete (== dual carriageway), and
> mostly limited access with infrequent at-grade intersections.  Urban
> areas are so crowded that roads that meet this definition have to be
> basically motorway like but probably more curvy with lanes that aren't
> wide enough, and have too many on/offramps.

This is illustrative of the problem we have in OSM of so many tagging
norms. Because this _isn't_ the way trunk is used in many other
countries. In Ireland, we tag our roads by and large by administrative
class, not by quality (though there is quite good correlation between
the two). For this reason, our highway=trunk sections can be anything
from motorway-grade roads down to quite poor single-carriageways. They
can also be surface roads in towns.

I know that German tagging norms have overloaded the interpretation of
trunk, something I was never happy with. Annoyingly, the German
mappers can't seem to agree on exactly what extra must-haves apply to
a highway=trunk. Some claim it should imply motor_road=yes, others
insist on grade-separated junctions, others still on near-motorway
characteristics.

My point here is that we need to be very careful before we assume any
consensus or even any possibility for real consensus. We have already
diverged on country lines in our interpretations of the highway tag
and it seems likely that any routing application will need to infer
quite a bit based on knowledge of local tagging norms.

If consensus is to be pursued, I am strongly in favour of the existing
highway tag being used to convey what I'm going to call road "class",
by which I mean its official classification, importance, routing
significance, whatever. Other aspects related to the quality of the
road or its suitability for any particular kind of vehicle or user can
readily be accommodated using other tags, some of which are already
widely used. We have, for instance, tagging norms for:

Prohibitions on use
Max and min speeds
height, width or weight restrictions
Number of lanes
Width of carriageway
whether single or dual-carriageway (not through tagging, but we can tell)
Grade separation or otherwise of junctions (through examination of
whether single or dual-carriageway and evaluation of turn
restrictions)
Surface
And even smoothness, for those who require it

So I urge:

* DON'T blindly adopt a tagging idiom from another language just
because they have been more specific
* DON'T adopt a change to tagging norms that will break a lot of existing work
* DON'T have sneaky edits to the wiki that don't reflect actual consensus
* DO look to evolve current practice in safe, sane steps towards
something that solves real problems with current practice
* DO document on the wiki the kinds of problems we are actually trying
to solve (hard to have a useful discussion without this)

Dermot

-- 
--------------------------------------
Iren sind menschlich