[OSM-talk] Edit war on the wiki "map features"

Dave Stubbs osm.list at randomjunk.co.uk
Mon Dec 1 11:08:33 GMT 2008


2008/12/1 Robert Vollmert <rvollmert-lists at gmx.net>:
> 2008/12/1 Richard Fairhurst <richard at systemed.net>
>> Personally I believe the easiest and most flexible thing is just to
>> extend
>> the access tags:
>>
>> bicycle=no|yes|difficult|unsuitable
>>
>> so you'd get
>>
>> highway=bridleway
>> foot=yes (permitted, no problem)
>> bicycle:racer=unsuitable (permitted but not practical)
>> bicycle:hybrid=difficult (permitted but challenging)
>> bicycle:mtb=yes (permitted, no problem)
>
> The obvious problem with this is the massive redundancy. You need to tag
> for every possible form of transport, or infer suitability for something
> exotic from the provided suitabilities.
>
> On Dec 1, 2008, at 11:09, Douglas Furlong wrote:
>> This feels like a far more suitable solution, than smoothness (and
>> Ice rink is smooth, but I doubt a racing bike would have much fun on
>> it!).
>
> Hurray for absurd arguments. Obviously, 'slippery=yes' is implied on
> ice rinks.
>
> I do wonder why people are always jumping on the corner cases to
> discredit
> smoothness=*. Would one of you that think smoothness is worse than
> nothing
> care to comment on the "definition by example" I proposed in
> http://lists.openstreetmap.org/pipermail/talk/2008-November/031779.html
> ?


Because the corner cases sometimes express the fundamental problem in
an obvious unsubtle way. Sometimes people point out corner cases just
to be annoying :-)

The reason I think no-one has commented on your definitions is because
they're infinitely better than what was actually being proposed, and
also have virtually nothing to do with what was proposed (they
actually seem to be related to smoothness!). The reason for that is
they describe a number of fairly specific examples, reduce the number
of options, and don't even try to classify the "rougher" examples
which is where most of the contradictions kick in. ie: a couple of
tree roots on a cycle path make it good, lots make it intermediate, no
mention of tanks.

The question most people will come up with is, what do I classify a
cycle path with an intermediate number of bumps as? You'd probably
declare it doesn't matter, which I'd be inclined to agree with. The
worst that happens is that a way spends the rest of its existence
oscillating between the two states. You could probably relate it in
some way to the cat/toast paradox, and if you could figure out a way
to generate electricity from it you'd become rich.

Or introduce a deliberately fuzzy border value like bad/intermediate.
And wait the necessary 20 mins for someone to point out this is
qualitatively different from intermediate/bad.

Anyway, the main problem with definition by example is what happens
when the examples don't fit. This happens surprisingly often. You
basically have two choices, just guess, or introduce a new example for
everyone to argue over.

You're also only defining the value for a specific purpose, which is
generally less useful than objective features which can be used for
many purposes. Of course if you're only interested in the specific
purpose that distinction isn't going to interest you, which is fair
enough.

Dave




More information about the talk mailing list