There was a debate about mapping railway=stations as nodes vs. areas in October 2020 which I would like to continue. The discussion started when @DaveF changed the old tagging scheme illustration on the Wiki page to the new, current one.
The old scheme
featured public_transport=station and public_transport=platform tags
advised mapping railway=station to cover the area of the whole railway station as defined by railway regulations (between entry signals on both sides of the station).
The new scheme
doesn’t feature public_transport=station and public_transport=platform tags
advises mapping railway=station to cover the same area as public_transport=station.
After a back-and-forth of @dieterdreist undoing and @DaveF redoing this edit, a discussion was (re)opened in the Tagging mailing list by @dieterdreist asking for further comments—hence the warning template on the Wiki page.
@aharvey voiced support for retaining the old tagging scheme but said that “maybe a solution is keep railway=station and public_transport=station both defined as the passenger view, but use a new tag for rail infrastructure so you can still correctly map the station for train drivers.”
@Mateusz_Konieczny mentioned that “I tried mapping some railway station as areas and I ended not doing this. Either mapping would be quite arbitrary or include massive area that is not really relevant.”
@dieterdreist stated that “tagging concepts should accommodate both, the general mappers and the experts”.
(There was also a discussion about this in the Thailand community started in April 2020—the author decided on mapping railway=stations as nodes in the end.)
My opinions on these, respectively:
I think there is no need to have two separate tags for having stations mapped just from passengers’ point of view, as they would result in duplicates and create confusion.
I understand that it would be a lot of work to map greater stations. But first, smaller (4 or less tracks) stations make up for the vast majority of railway stations, where mapping would be easier. Second, it is indeed relevant to have an area for the whole station – as defined by railway regulations, which in Hungary is between entry signals on both sides of the station – because otherwise you can’t pair most railway station infrastructure elements (side tracks, switches, signals etc.) with the station they are located in. It would be really important to have these connections, e.g. I have a project where I need to download all infrastructure elements of a station from OSM, but I can’t do that now, because there is no easy way to tell which side track or switch belongs to which station. So I think it would worth it.
I agree that tagging should accommodate both general mappers and experts. In my opinion public_transport=station (which is mostly used on nodes) is suitable for passengers and general mappers. But I think tagging schemes should also enable expert use cases like the one I mentioned in the previous point.
Therefore I would recommend restoring the old tagging scheme on the Wiki page and adding a recommendation to map railway=stations as areas.
I apologise for the lengthy post, and I’m looking forward to your thoughts on this!
I think there is no need to have two separate tags for having stations mapped just from passengers’ point of view, as they would result in duplicates and create confusion.
there are also stations that are for goods (and no public access)
A railway=station in North America, in the infrastructure sense, can be everything from a sign in the ground trackside (named station in “dark” territory without a siding) to a very complex area containing multiple interlockings. Also, passenger stations here may or may not have any association with signaling infrastructure, whether they be tiny halts or surprisingly significant stations on S-Bahn-trunk-ish setups. Furthermore, there may not be anything demarcating the “on the ground” boundaries of some stations – industrial and yard tracks will be associated to a station in common practice, but which track goes with which station can be quite complex in the context of a large yard.
railway=halt for stations without switches is common practice in some countries like Germany, but there isn’t global consensus on the distinction between railway=halt and railway=station. For us in North America, a halt is a passenger station with flag-stop service (i.e. trains stop on request only).
Maybe it would be a good idea to mention on the Wiki page that if station boundaries can’t be easily defined, map it as a node, but as an area otherwise.
Thanks for pointing that out! I would also recommend mapping railway=halt as an area (if halt boundaries can be defined) if the halt has switches for the same reason:
it implies one train cannot pass another one, so this is functionally a significant distinction, but I agree it is not necessarily the distinction made between halt and station
The presence of points/switches at a station does not necessarily mean that trains can meet/pass there. An industrial track freight station in the US with a non-clearing spur (i.e. a spur off a fast CTC mainline that lacks leaving signal or electric lock protection) would be an example of a place where you can’t have trains pass despite having a switch and a track to do it with.
Indeed, I find it very odd that Punt Muragl Staz is a different kind of station from Celerina Staz. The latter has a small siding, but no ability for trains to pass.
Many other RhB stations are unattended request stops, as do numerous stops on lines in the Scottish Highlands, which conform reasonably with original railway usage, are all tagged railway=station. As presence or absence of switches (or branching of railway tracks) can be ascertained with relatively simple queries, it looks as though railway=halt is not a particularly useful tag.
Until recently my local station was Furze Platt, which was, and largely still is, a classic halt in both the original UK sense, and the DACH OSM sense. Indeed it was originally called Furze Platt Halt (only renamed in 1969). The fly in the ointment is that it has a lot of passengers (~180k pre-covid), and consequently also has a ticket office open in the morning weekday rush hour.
im deutschen Forum gab es vor fast 10 Jahren die gleiche Diskussion und soweit ich verstanden habe, begann dies mit der gleichen Idee, aber das Ergebnis war dann ein anderes.
Another good corner case to consider is the Amtrak station at Winter Park – there is a turnout there (East end of Winter Park siding), but it’s not part of the Amtrak station! (The infrastructure owner on that line has their own notion of a “station” at Winter Park that has nothing to do with the Amtrak station there.)
Thank you very much for pointing it out, I was not aware of that discussion!
The decision there was the following:
To sum that up in English, if I understand it correctly:
Stations should remain tagged as railway=station nodes
The station area (defined by local railway regulations) should be tagged with landuse=railway
This area should be linked to a public_transport=stop_area relation.
Objects tagged with public_transport=platform, railway=stop, railway=subway_entrance, landuse=railway, building=train_station and railway=station or railway=halt should be added as members to this relation.