On Sun, May 24, 2020 at 8:01 PM John Willis via Tagging
<tagging@openstreetmap.org> wrote:
> Mapping “where the sidewalk ends” and the trails begin is vital to keep
> people from being routes where grandma could have a heart attack Climbing a
> difficult route or break her leg crossing a stream because we routed her on a
> trail down a ravine rather than on the longer, yet safer sidewalks down along
> the roads or paths through a local park because there is no way to say “THIS
> IS A TRAIL, not a walkway through a playground” in OSM.
We do have that: `sac_scale=hiking`; as I understand it, few trails go
beyond 'hiking', so that's at least better than nothing. (It still may
suffer from underestimating the trail, leading city folk to the
sketchy rock scrambles when they're expecting a nice level dirt path,
so try to get the scale at least reasonable.)
What we don't have - at all - is the complement: 'THIS IS INDEED A
PATH'. When we see 'highway=path', we don't know whether it's indeed
a path, or a hiking trail where the mapper didn't assign an
`sac_scale`. We need a way to assert 'THIS IS A PATH' that doesn't
depend on the absence of a trolltag.
I can't stress enough that as long as we have the ambiguity, the only
way to 'fail soft' is to support the assertion 'this is relatively
safe', because we can deduce nothing from the absence of a 'this is
dangerous' assertion.
Incomplete information should 'fail soft'.
On Sun, May 24, 2020 at 8:32 PM Andrew Harvey <andrew.harv...@gmail.com> wrote:
> Agreed, the biggest question is how do you define that criteria for what is
> going to be tagged a a hiking trail and not a hiking trail.
>
> Eg. if you have a smooth paved track through the rainforest that the
> authorities created for grandparents and strollers, is that a hiking trail
> just because it's in a forest area? What about a stroll through the hills of
> grasslands that have no forest or mountains, is that marked as a hiking trail?
No, just being in a forest doesn't make something a trail. I think
that it's pretty safe to assume that 'surface=compacted
smoothness=intermediate wheelchair=yes` with a connection to a highway
or parking area not strictly a hiking trail, and there are some trails
near me- even in Wild Forest areas- that are constructed in such a way
to offer wildland access to persons with disabilities. I'd be happy
considering those trails on an equal footing with urban paths.
A hiking trail can be an easy trail through the lowlands. (Those are
rare near me, because the lowlands are mostly either settled and
subdivided, or else sucking swamp, so the mountains are what is left
for hiking trails to go.) I already mentioned that sac_scale discounts
hazards other than mountains (and focused on water, but Graeme can
surely fill in a number of deadlies that are specific to his
continent).
A lot of it comes down to, "would you route mobility-impaired people
or folks with small children in tow down this?" A wrong decision for
some ambiguous corner case will be mostly harmless. Not having the
information for a dangerous trail might be deadly.
> I think it's too hard to have a reliable criteria for this which can be
> objectively surveyed, it's much easier to tag each attribute individually on
> their own independent scale.
The _reductio ad absurdum_: by the same token, because there is
controversy in many locales over which highways should be
`highway=trunk` and which should be `highway=primary`, or which should
be `highway=service` and which should be `highway=track`, all highways
should be tagged just `highway=road` and the relevant attributes
(surface, smoothness, speed limit, number of lanes, ...) should be
mapped instead. Few if any of us think that would be appropriate. Why
can cars get a hierarchy of ways, while hikers, equestrians and
cyclists cannot?
Since the possible set of attributes is open-ended, the result of not
having some sort of categorization is a nightmare for data consumers,
trying to determine how to render a road, or whether the road is
routable in the current circumstances, or where there are 'trails for
hiking near here'. No sensible symbology can map all the possible
attributes, and no sensible router can take all of them into account.
At some point, _someone_ has to make the call of what is considered
suitable, and punting that decision all the way to the end user is
what leads to the sort of accidents that Graeme, John and I have been
discussing.
Even an 'objective' attribute often turns into a controversial
position; consider 'car=no' versus 'car=private'. A lot of mappers
think that 'no' should be reserved for ways on which it's physically
impossible to drive a car: What about the corner case of ways that
could accept only a car with high ground clearance, or a way that a
skilled stunt driver could manage but most drivers could not? So now
we need to come up with separate tagging indicating the legal status,
versus the physical status, and a detailed physical model of how the
way affects the handling characteristics of an automobile? And even
after you reduce everything to physical numbers, you still have the
corner cases where the precision of the available measurement doesn't
support assigning a category unambiguously ('This can't be a track
because it's only 2.499 metres wide, and we require 2.5.' 'Uhm, I
measured it by pacing, I doubt my figure is right to within even a
centimetre!'). Where do you stop?
As I said, the possibility of getting the classification wrong in some
equivocal corner case is pretty harmless - much less worrisome than
having no way to make the classification short of negotiating a maze
of attributes.
The map is not the territory; it is, however, a model of the
territory. To quote George Box: "All models are wrong. Some models are
useful." Tagging should be objective, but no matter how hard we try,
we cannot remove the subject entirely from the object. I'm loath to
dismiss a tag as 'subjective' merely because there's room for judgment
at the edges.
(See also piste:difficulty - addressing safety in a different sport.
Virtually all ski resorts everywhere use some version of the
green/blue/red/black/double-black-or-orange system and I haven't heard
anyone complain that it's 'too subjective'.)
--
73 de ke9tv/2, Kevin
_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging
