On 21/09/2014, Tod Fitch <t...@fitchdesign.com> wrote:
> Despite being actively discouraged, "paved=yes/no" is
> used. And two of the top values for "surface=*" are "paved" and "unpaved",

A lot of those are historical values, before the practice of distinct
surface values took hold.

> clearly taggers find the concept of "is paved" versus "is not paved" a
> natural one. And I strongly suspect you would get a more consistent result
> from an arbitrary person trying to "map what you see" if you asked them to
> look at a road and determine if it was paved or not than if you asked them
> to specify the name of the surface material.

It would be nice if it was true, but it isn't. Consider
surface=compacted : while mappers do have a clear idea of what is
"paved" or not, that's the kind of surface thay'll yield
random/subjective paved=yes/no answers. Or consider
surface=cobblestones : while everybody would tag that paved=yes, a lot
of data users who look for "nicer" roads will want to avoid that
particular kind of paved=yes. They're just two examples amongs many
that show that a binary value is not as interesting as it sounds.

As a user, I'd avoid a router that only cares about paved=yes/no.
Looking at surface=* instead isn't hard. You can probaly afford to
just look at the ~30 most common values (ignoring typos and rare
items) and still get less issues than you'd get by looking at
paved=yes/no. As an added bonus, you can make your own selection of
what surface is "nice" for your usecase, and even use nuanced ratings.

Tagging mailing list

Reply via email to