[OSM-talk] General tagging strategy
osm at tobias-knerr.de
Wed Jul 9 21:05:38 BST 2008
> Strategy 1: Legal status. (e.g. there's sign saying it's a cycleway)
This is the most useful and most important one. It should be easy to
find out, and it is objective information. It is also what people expect
to find on their maps. The other things will probably only appear on
specialist maps or be used for routing.
> Strategy 2: Common usage. (e.g. people are riding their bikes there,
> even if it may be forbidden -or- even though it's allowed to ride a bike
> here, it is a really bad (suicidal) idea to do so)
That’s the worst strategy in my opinion, because common usage depends
on 1 and 3/4 – why would people cycle illegally on unusable roads, and
if they do, why should I care? It’s also rather subjective information,
unless you actually provide some statistics.
> Strategy 3: Physical properties (e.g. gravel with a average corn size of
> 1/8th inch and an average of 3 bump holes between 2 and 4 inches in
> diameter and 1 to 2 inches deep per meter)
In theory, this would be #2 on my list, because legal status and
physical properties determine usability for all types of transport.
(Which means that we could provide usability information for small data
user communities. Imagine, for example, that we manage to get many
cyclists enter data about cycleability into OSM. We then can provide
good routing for cyclists, but have only vague estimates for inline
skaters – unless the cyclists have provided us with actual physical
properties from which the information about usability for inline skating
can be derived).
That being said, it is – unfortunately – extremely hard to
a) find good tags that allow to enter all the relevant information
b) enable software (routing apps etc.) to use the gravel corn size and
bump holes in their algorithms and generate usable results in the process
c) collect the data.
Still, if someone succeeds at a), I’d encourage c).
> Strategy 4: Interpreted usability (e.g. usability for bikes is 2 on a
> scale from 1 to 5)
Subjective, unpleasant from a modeling view (different kinds of
information all stuffed into one, almost as bad as that highway tag) –
but the only realistic way to get appropriate routing in all but very
long terms. We’ll have to go for this, I guess. (We might want to
discuss the numbers vs. descriptive values issue, though.) It won’t stop
us from using the highly detailed strategy 3 data where it is available.
In short, enter data from categories 1, 3, 4, with priority on 1.
More information about the talk