[OSM-talk] The Return of the Highway tags and other junk

[David Earl]

Ben cited an example...
> highway/railway=viaduct
> highway=gate,
> highway=cattle_grid
> highway=footway
> highway=track
> railway=abandoned
> railway=narrow_gauge,
> foot=permissive

This isn't really a direct reply to your proposal, more a response to your
example which raises a number of points of principle which are interesting.

Gate and cattle_grid go on their respective nodes, not ways or segements, as
I understand the spec, so that's not a problem is it?

And surely a footway and a track are mutually exclusive by the definitions
in the spec. I wasn't party to the original thinking, but I presume the
assumption is that in general there is a hierarchy so that one can walk
along a track (or a cycleway or a primary) unless there are access tags on
it, except for well-understood special cases such as motorway (or e.g.
one-way assumed on a roundabout). Track means 'also a footway' by
definition, yes?

JOSM doesn't let you create them, but is there anything fundamental to stop
one having multiple tag/value pairs with the same tag? The XML
representation doesn't prevent this, and unless the 'way,tag' pair is
defined as a unique key in the database, nor does the database table.

'abandoned' and 'narrow_gauge' are certainly both independent properties of
a railway and there is a flaw there if one cannot have multiple instances fo
the same tag. The same applies to the route tag - a highway can be a carrier
for many routes (e.g. a national cycle network route, a European route, two
trunk routes and so on).

Slightly away from the original topic, but when is an abandoned railway no
longer a railway? If it is reused as a track, has it then disappeared as a
railway, it's use subsumed? Taking this further, is an abandoned raiway
still such if it has housing built over it and is no longer a visible
feature in the landscape?

'Course of' railway (or Roman road, or filled in canal) may be useful as
historical information, but that's different from a visible bit of the
landscape. A bridge over the old railway has become a bridge over a
cycleway, even though it's historical purpose was to serve the railway. So
maybe the answer is to mark these independently but in the same place. The
fact that they follow the same course over some of their length is not
coincidence, but they are different entities. This way the course of the
railway can continue on the map under the houses built across it. Marking
the old railway has the advantage that it explains some of the features
alongside it (cuttings, embankments, viaducts), but while those still exist,
the railway itself does not.

To take an example where two features coincide and both definitely still
exist, in Wuppertal in Germany the mass transit system (dating back to
1901!) is very imaginatively built as a monorail hanging over and following
the course of the river for much of its length
 http://ktransit.com/transit/Germany/Wuppertal/Photos/wup-mr13.jpg
How should this be represented?

I'd assume that these should be two independent Ways following the same
segments (or perhaps even separate segments and nodes that happen to be
superimposed). Doesn't the same apply to an abandoned railway and the track
which it's course is now used as? (Thinking about it there are lots of
examples like this - the Manhattan Bridge in New York, has two decks one
above the other, subway and road)

Is the viaduct then a separate independent feature, which is for reasons of
convenience attached to a Way, but if one were being pedantic should be
separately represented? (What happens when a viaduct carries two adjacent
Ways? The most common example is a dual carriageway - two separate ways in
the spec - are carried across the same physical structure in some cases, but
often there are separate bridges for each carriageway. Do we care? Is it
worth the bother? At the moment we can't represent the more common case
where a they share the same structure because of the way the carriageways
are represented).

If we make it too complicated, the cost of creating the thing puts people
off. If we make it too simple there are odd cases like Wuppertal or shared
bridges or new features following the course of an old railway, which are
anomalies. In the end does it matter much?

If a change is proposed which can't be (mostly) automatically converted from
existing data, it needs a great deal more justification IMO than one which
simply adds a new kind of feature. Manual changes at this stage are very
hard to achieve, and it is probably better to live with a flaw in the
original spec.

David