[OSM-talk] Proliferation of path vs. footway

[Ben Laenen]

Greg Troxel wrote:
> John Smith <delta_foxtrot at yahoo.com> writes:
> > It's most likely going to have to be jurisdiction specific, not just
> > country specific in some instances. Going the other way and dealing
> > with footway for example, NSW & Vic doesn't allow cyclists on
> > footpaths, but ACT does.
>
> This is a major philosophical issue for OSM.  I think it would be really
> nice if there were a consistent scheme so that one could interpret what
> was reasonable/allowed from tags alone without needing to know which way
> local defaults to.  To some extent we are relying on similar traffic
> laws in enough countries to declare a default, and to expect places
> where those defaults are not true to tag as exceptions.

Except it's plainly impossible that someone is going to be able to make 
defaults for the entire world, and certainly no defaults that the entire world 
would be happy with. A large number of tags are already dependent on the 
country they're used in anyway. So trying to come up with global defaults 
doesn't make sense anymore, and will certainly make people in some countries 
unhappy if they suddenly have to add several tags to each way where those tags 
used to be implied, because suddenly global defaults were introduced.

And global defaults immediately break down at step one: defining vehicle 
classes. Those are just different in each country, and simply can't be merged.

The concern is also that:
(a) traffic code changes over time, so what's allowed today maybe wouldn't 
next year, even though the signs would stay the same, so we have to map in a 
way that we don't have to retag all roads if it does change (or worse: revisit 
each road because we don't have a method to know what was on the road)
(b) mappers don't exactly know the entire traffic code either so they wouldn't 
know of certain exceptions or rules, they can see the traffic signs though so 
it makes sense to have tags that somehow relate to those signs in an 
understandable way

I'm not saying we have to use traffic_sign=C9:M2 or something, but a tag that 
more or less says what's on the sign (like motorcar=no when it's a prohibition 
sign with the icon of a car), even when it's not just motorcars that are 
prohibited there, but all motorized vehicles with more than two wheels, like 
motorcycles with a sidecar.


> In the case of cycleway, the global norm seems to be that pedestrians
> are permitted.  But we have to do either
>
>     define a default for each jurisdiction
>   AND
>       encode the default in the map with polygons
>     OR
>       have some table for renderers)
>
> OR
>
>     define a global default
>   AND
>     tag all ways with exceptions to the global default

It definitely has to be
    define a default for each jurisdiction
 AND
    have some "table" (a library or whatever) for programs that need to know
    the rules

But making that "table" will certainly not be an easy task... And table may be 
a bad word here, it would probably have to be something rule based, with for 
example conditional tests on certain tags, as one tag won't be enough to know 
access rules of all vehicles.


> With highway=path, the wiki page does not give the semantics when there
> are no tags.  For highway=path and no tags, is that horse=yes or
> horse=no?

I can only say how it's for Belgium: on a highway=path these vehicles are 
allowed unless there are other tags:
* foot
* bicycle
* horse
* moped_A
* moped_B
(those last two are subclasses of moped)

> Is it paved or not if there is no tag?

It's just unknown if there's no tag. Is a default really needed for surface?

> The biggest problem is that there needs to be an unambiguous mapping
> From these highway=foo tags to the implied value of the access subtags.
> The next biggest is non-operational semi-circular definitions like
> 'highway=cycleway' being for 'designated cycleways' which talk about
> 'intent', although in practice one would ask (in en_US) "do most people
> think this is a bike path".

Can't agree more. I've done some work towards it for Belgium here 
http://wiki.openstreetmap.org/wiki/User:Eimai/Belgian_Roads#Paths
Note how cycleway/footway/bridleway are only used for paths with the round 
blue signs.

> Maybe we'll end up with a definition of bridleway, cycleway and footway
> in terms of path, with the notion that cycleway and footway are paved,
> and path is unpaved if not otherwise tagged.

There's really no need to define cycle/foot/bridleway in terms of path. They 
can have their own special use. Many legislations will have their notion of a 
cycleway, footway or bridleway, and they'll have their implications in terms 
of access rules (not necessarily on the path only, but as well on the adjacent 
road for example), who has to give way to whom etc. So it makes sense to align 
those definitions with the usage in OSM.

Greetings
Ben