[Tagging] New key proposal - paved=yes/no

[Tom Pfeifer]

-1, because:

Tomasz Kaźmierczak wrote on 2014-09-20 23:42:
> I would like to suggest making the paved key for highways (and probably other types of elements) official.
> Taginfo for paved:
> http://taginfo.openstreetmap.org/keys/paved#values
>
> The above shows that the key is already being used,

1648 times, compared to 9383813 of 'surface'.

> but the Wiki doesn't describe this key, instead redirecting Key:paved
> to the article about Key:surface.

to discourage the use of a duplicate key

> However, in some cases, a simple information on just whether a highway is paved or not, would be very useful.
> One such case would be navigation software – if a user chooses to avoid unpaved roads,

Navigation software is pretty able to consider a short list of specific pavings
as 'paved' and another short list as 'unpaved', they are already structured in the
wiki.

OsmAnd, as a popular navigation software, does so, and in the pre-1.9 nightlies you
can switch on colour coding for different surfaces.

> the software can check the value of the surface key, but in practice most (all?)
> of the navigation software only checks for a subset of all the possible values
> the surface key can have.

Could you please support your argument with examples of such software, and
why such incompleteness cannot be fixed within the router/renderer?

> If the paved key was widely used, then the navigation software would have a simple
> and clear way of checking whether a given road is paved or not.

Not much simpler than checking for a member of a list.

> The default value
> of the paved key for highways could be yes, so that it would be consistent with the
> assumption that highways in general are paved.

This does not work as a general assumption.
I would assume a motorway as paved, but a track or path as unpaved, unless shown otherwise.

> Also, the surface=paved could also implicate paved=yes
> and similarly surface=unpaved could implicate paved=no, so that duplication of the
> information could be avoided when the generic paved and unpaved values are set for the surface key.

You are just arguing against your proposal. As we have surface=paved
we don't need paved=yes. And surface=asphalt implies paved.

Tod Fitch wrote on 2014-09-21 01:15:
 > It might be considered duplicative, but what should a data consumer do if confronted
 > with a surface=* value that is unknown to it (and the wiki)? We aren't talking human
 > intelligence here where an informed guess is possible. Should such a way be considered
 > paved or unpaved for purposes of routing? From that point of view, Richard's proposal gives a lot of clarity.

You would not want to add 9383813 unnecessary paved=yes/no just to cover
a few dozen undocumented values for surface, most of them probably typos.