kevin.b.kenny+osm at gmail.com
Sun Oct 8 20:12:08 UTC 2017
On Sun, Oct 8, 2017 at 1:28 AM, Dave Swarthout <daveswarthout at gmail.com> wrote:
> I m following this conversation in hopes that if it ever gets resolved
> someone will update the Wiki. I have my fears that, along with many other
> contentious issues, it may never be resolved to the satisfaction of all
> Meanwhile, I'm doing major work in Alaska and although my current focus is
> primarily on adding geographic features, this issue has practical
> implications for me. The George Parks Highway and the Alaska Highway come to
> mind immediately. They are a bit of a mish-mash with some sections tagged
> motorway, some trunk, and the speed limit varies from 65 mph in rural areas
> down to 40 mph in towns. That's the nature of the highway system in Alaska
> where a single highway serves an immense largely unpopulated geographical
> and area. Most sections of those highways are "trunk" roads by most
> definitions yet they have normal at-grade intersections, intersections with
> driveways, tracks, etc.
Personally, I would not tag those two highways with anything below
'trunk', no matter how bad they get. (I can surely remember a time when the
Alaska Highway was not hard-surfaced owing to the difficulties of laying
down pavement over permafrost.) They are of too great regional importance
to show as anything less.
And this discussion is offering me, at least, some insight that may be
obvious to everyone else. The mention that there are three things being
conflated: administrative level, regional importance, and physical character;
has set some of the mental wheels in motion.
Looking through the lens of a data consumer:
Administrative level, which NE2 confused with the other two, is pretty
much a 'don't care.' If a highway is a freeway, for drawing it on a map,
it doesn't much matter whether it's Interstate 95 or County Road 40. All
that really effects is the shape of the shield on the highway's placards.
Relative local importance is what a renderer will use to decide whether
or not to show a road at a given map scale. I'd be entirely comfortable if
a map that fit all of the Northeast on a single piece of paper were to omit
a lot of the suburban motorways. I would not appreciate it, however,
if such a map in northern New England were to omit US highways
1, 2, 3, 4, 7, 201, 302; Maine 11, 16, 27, and so on. They are just
two-lane roads, often with lowered speed limits and narrow shoulders,
but they are the arteries of the communities they serve. You can't
get to Rutland (the third-largest city in Vermont) without driving
one of those trunk roads - if trunk roads they are.
Physical character is what's of most interest to routers - the speed at
which you can traverse a road and the volume of traffic that it can
handle are the key things that a router will want to have in order to
decide whether to send traffic there or elsewhere.
Similarly, physical character is really what a renderer should use
to choose HOW to render a road (rather than WHICH roads to
render at a given scale. A freeway remains a freeway even if
it is of only local importance. A narrow, winding, two-lane
highway remains such even if you have the misfortune of
needing to drive a hundred miles on it to reach your destination.
So, all three attributes are important. (Administrative level is
the least so, except as a poor surrogate for 'regional importance.')
In a well-ordered country like much of Europe, administratative
level is a much better indicator of importance, and important
roads are more reliably maintained to a higher standard, so
all three attributes tend to run quite closely together. We
Americans inhabit a country that is fundamentally messier,
and this has led me to tag combinations like "highway=secondary
surface=compacted smoothness=bad lanes=2" for particularly
bad roads that are still the only reasonable connection between
two populated places.
Perhaps we could reach consensus more easily if we were
to first try to agree that the goal is to tag both physical character
and regional importance, and recognize that the two serve
different needs, and are (in the US) often grossly mismatched?
Then the discussion could revolve around the question of what
tagging is for physical character, what tagging is for regional
significance, and what are objective criteria for assessing
significance. (It's somewhat subjective, and therefore
contrary to the OSM spirit of "tag what is visible only on the
ground", but it's so necessary to getting mapping and routing
right that I think we have to grasp that particular bull by
More information about the Talk-us