On 18/05/2011 12:22, Andy Allan wrote:
On Wed, May 18, 2011 at 11:03 AM, Richard Fairhurst
<[email protected]> wrote:
That way, renderers and routers could choose
not to show refs which aren't helpful for their audience. Something like
ref:signed=no would work.
Any thoughts?
Personally, if it's not signed, known, called or otherwise indicated
in any way to have that reference, whether in atlases, satnavs, online
maps, streetsigns or anywhere else, I'd rather we didn't use the "ref"
tag for it. If there's a special super-secret special reference only
used by highways authorities and nobody else, then it should go in a
(super-secret) special tag.
Otherwise we're back to the "this road has reference X, oh no it
doesn't" school of tagging, which we've agreed in the past isn't
helpful.
I agree the "ref" tag as it is is rather ambigouous. As I (would like
to) read it, the "ref" tag (not just for roads, but for many objects)
should really be the "unique identifier" by which it is known in the
administration of the responsible authority. There are many, many cases
in which the highways authority has a different "ref" to how it is
visible on the ground, or known locally. Think of motorway link roads:
the link is often administratively part of the motorway it is coming
FROM, whereas the overhead signs lead you to think it is part of the
motorway it is leading TO. (This is the way it is in NL where I am now).
I have been thinking recently about a way of reflecting this in the
tagging and removing the tension between the two possible definitions of
"ref". My idea at the moment is that the existing "ref" tag be
unambiguously defined to be a kind of "primary key" and to have
alternative tagging for its apparent value. I expect this may also be
the case with road names, and possibly other attributes as well where
there is an opportunity for the official version to deviate from the
apparent/published version. How about this for a hypothetical link road
from the M1 to the M2:
ref=M1
signed_as:ref=M2
In the case of C-roads, we could have something like:
ref=C123
signed_as:ref= (i.e. blank, or a special value like "no",
unless it actually appears on the signs!)
General purpose renderers (including mkgmap) would give precedence to
the signed_as values for ref and name if they exist. This would make the
resulting maps far more logical to end users, while allowing the
official version to have a proper home in OSM for those renderers/users
who want it.
Any thoughts on this?
Colin
_______________________________________________
Talk-GB mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/talk-gb
