[Tagging] Feature Proposal - RFC - trafficability

Ronnie Soak chaoschaos0909 at googlemail.com
Sat Jan 18 09:03:39 UTC 2014


I would also tag these things as free text.
The problem space is just too big to encode this in standard tags with a
fixed set of values.
If you can express the problem more specific and even with less words in
free text then  in key-value pairs, I would clearly vote for the former.
Especially as a lot of those tags would be quite unique and lack data
consumer support anyway.f

It is just not possible to tag e.g. the whole of Iceland in a way so that
you can reliably say which roads are passable by you vehicle and which not.
Even when you flood the database with 40+ tags per highway (which makes
them almost unmaintainable by mere humans), input all your vehicle
information into the router (which will take you the whole day) and there
is a precise
machine readable weather forcast, there are STILL factors you can't know,
that change on a daily basis or depend in chance alone (5cm of water level
on a river crossing can make the difference). And even one of those will
mean that your calculated route could change completely.
You may get a 'forecast' that's about 50% sure to the cost of having almost
destroyed OSM for manual editing.

If you really think routing based on chances is a goal worth pursuing, than
you may also want to tag just that. chance_of_passability = 34%. Yes, this
is a completly subjective value based on no deterministic algorithm (and
therefore start edit-wars), but the effect on routing will be exactly the
same while being easy to edit and maintain. And in the end, that is exactly
what you get when some local tells you that the road might be passable or
not. You can ask your router to give you a route with a risk (= 100% -
combined chance) less than e.g. 2% and you will be routed on primary roads,
accept e.g. 20% and you might fail when it's muddy, accept 90% and you
might only pass on best of conditioins if you come well prepared. Again:
imho even with a myriad of different special tags, in the big picture the
reliability of routing will not get much better than this. So why bother?

If you really really want to be more precise, I think weather conditions
have a value set that is almost managable in size. So
chance_of_passability:wet = 10%; chance_of_passability:dry = 80% might be
possible. with a fixed (and therefore machine-evaluatable) set of subkeys.
But imho you can never achive this for vehicle specifications or physical
road condition. The set of possibilities is just to large.

A free text note might help to judge the chance a little bit better and is
easy to maintain and render. So: why not? If translation is the only
problem, I think we might be able to handle this over time.

my 2 (more of 200) cents

Chaos
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20140118/182eee91/attachment-0001.html>


More information about the Tagging mailing list