[Tagging] Feature Proposal - RFC - More Consistency in Railway Tagging

Martin Atkins mart at degeneration.co.uk
Sat Apr 13 18:11:32 UTC 2013

On 04/13/2013 10:18 AM, Rovastar wrote:
> Martin,
> The example you gave for tunnels and bridges are the same for roads as well.
> If you have a bridge or tunnel with 2 roads (one for each one-way) and a
> train line(s) and footpath each will be a tagged with a separate bridge. So
> in that regard rail is actually are consistent with the road network.

Point taken: it is a general problem with bridge tagging, not one with 
railway tagging. I found some discussions on the wiki about modelling 
bridges as areas that would address this, but I don't really have any 
interest in mapping bridges in particular, so example retracted for the 
purpose of this discussion.

> Also you say you want it better for simple mapping and other can do more
> detailed mapping if they want to. I see no part of your proposal to add
> additional tracks like is now yet you imply in the posts here that it is. If
> you do think this then it doesn't not help the crossing example you gave as
> we will have the same problem again.

I have not described a way to describe the actual routes of tracks; I 
lack the expertise (or interest, frankly) to describe that next level of 
detail, I just propose that we separate that next level of detail from 
this simpler level of detail, rather than using the same tags for both.

I've seen the area:highway proposal for mapping the detailed shape of 
streets, sidewalks and footways. My assumption was that this proposal 
could be extrapolated to include a similar model for railways, modeling 
the precise shapes of the trackbed the rails run along, the positions of 
the individual tracks within that trackbed, etc.

If I were trying to define such a thing my first thought would be to 
define a new way tag to mean "the exact path of a track" and use 
separate ways from the simple route network. e.g. railway:track=rail . 
I've not spent nearly as much time pulling that idea apart as I have my 
simple route-oriented proposal, so I'm sure someone who knows more about 
railways than I do could find examples where that doesn't apply, but 
it's a strawman to start with.

I could also compromise on making the schematic network be the thing 
that gets new tags, but I think it's tough to say whether it's better to 
suggest re-tagging detail work in dense areas where there are evidently 
lots of avid mappers at work (and the re-tagging could thus happen 
relatively quickly), or to suggest re-tagging the basic network in areas 
where there is less detail and there are fewer (or no) active mappers.

> Do you propose that we change *all* the currently mapped multi track rails
> to conform to your new standard?
> e.g. here there are hundreds of tracks/railways which IMHO accurately
> reflects what is on the ground.
> http://www.openstreetmap.org/?lat=51.47119&lon=-0.14847&zoom=15&layers=M

Under my proposal it wouldn't do any harm to leave existing detailed 
tagging in place where the railway doesn't cross the highway, since a 
railway way represents "one or more tracks", and that holds for the 
example you showed. Optionally one could add tracks=1 to the existing 
ways to make it very clear.

The converse is not true: if you define that a railway way represents 
exactly one track, then there's lots of work to do to turn miles of 
existing one-way-per-railway tagging into one-way-per-track.

When it comes to tramways and level crossings, whichever approach we 
take there are many counter-examples to be corrected, of course.

I would note that if we *did* adopt a separate tagging scheme for 
detailed mapping of tracks then at least one could simply re-tag the 
existing railway=rail as (e.g.) railway:track=rail and not destroy the 
existing detailed geometry. Of course, someone would have to draw in the 
basic route network too; I bet the data for that is somewhere buried in 
the OSM historical record, since a schematic view of the UK railway 
network was imported into OSM as a starting point many years ago and is 
still the basis of simple mapping in many rural areas, but admittedly I 
have no idea how or whether it could easily be recovered for situations 
like your Battersea example.

> However what I do agree with you is that the rail guidelines should be more
> detailed but I would go the other way with saying that all tracks should be
> mapped not less for complete mapping. That is a common way of doing things
> and going forward especially as we get more detailed mapping (it's slowly
> coming to the US ;))

I'd be fine with that as long as the result includes details about how 
to connect the road and rail networks in a clear, unambiguous way at all 
levels of detail. It is the lack of definition around these interactions 
that causes the most difficulty, I think.

London isn't a great example of the problem since it has many, many 
railways but very few situations where railways connect with highways at 

Croydon Tramlink is one counter-example, and I'd concede that someone 
has done great work in accurately mapping the path of it that I wouldn't 
want to destroy, but it is sadly completely disconnected from the 
highway network, and that is what I'd like to address.

More information about the Tagging mailing list