[OSM-talk] Abandoned Rails

Lauri Kytömaa lkytomaa at gmail.com
Thu Aug 27 12:06:11 UTC 2015

(I hadn't subscribed to this list, so the reply is to a seemingly
random message and not directly related to that)

I believe much of this recent discussion is happening because
there's a ... misconception that hasn't been addressed, and
the actual tags that have been mentioned suggest readers to
believe so. I believe I've mentioned this idea in the past, but I
feel compelled to have it included in the discussion so that
the arguments on either side refer to the same concepts.

When a way no longer is an intact railway (or railbed),
we don't want to claim it "is a railway" but rather "this was a
railway"; the railwayness becomes an attribute of what is,
i.e. "this row of trees and this embankment were for a railway
and part of the railbed"), when previously it described
"an object", i.e. "this is a railway and railbed". In short, they
shouldn't use the same *key* in the tag. I therefore

instead of changing the tag value to railway=dismantled,
it would be better if mappers changed the tag key to
"was:railway"="rail" following the method of lifecycle
prefixes (quotation marks only for added readability)


The prefix can be dismantled: or razed: or destroyed: or
removed: or was: or something else; I personally prefer
"was:" as it's applicable to a lot of other cases, and applies
to all that "were" something. At least that way the
information (that the real world object no longer is an object)
is stored in osm, in contrast to plain deleting which doesn't
tell anyone if it was deleted because the way was a
mapping mistake or just replaced by a better version. Even
if it's deleted later by someone who finds it impeding his
editing, it is still *possible* to extract that difference. With
a change of key, even those who blindly draw all ways with
railway=* (for example the humanitarian layer on osm.org
seems to do so) won't be drawing false features.

Reassembling straight-away-deleted railway ways and
figuring out which ways presented the last coherent state
of things requires manual work from everyone who wants
to see that information (think 30 years from now), but if the
real world removal is first tagged, the most manual part of
that work is already done, even if somebody later deletes
the ways. And it would be at least possible to
automatically watch for and store all those objects in OHM
or similar, but discard the "technical deletions".

At least in urban environments the small details that tell a
railway existed can remain for centuries: unusual colonnades,
loading platforms, fasteners on the walls, curved buildings
in an otherwise square road network etc., so the line
between "totally gone" and identifiable isn't a clear cut line;
why would it be paramount to delete stuff just when the
iron beams were lifted, or when new asphalt was poured

A possible life of a railway section in urban environment,
a simple case:
- the railway construction starts: railway=construction
- the railway is in active use: railway=rail
 (say, in this example, in the middle of the city harbour, between
warehouses which even have loading platforms at the height
of the freight carriage floors)
- the railway is no longer needed, no trains run there:
 railway=disused  (everybody sees it's a railway). (a road had
been built between the warehouses)
- the track is converted to a sidewalk (the harbour is scaling
down), but the loading platforms and the geometry remains;
the way was a railway, and can be identified as such with
expertice, and/or local knowledge and/or old sources: following
the method of lifecycle prefixes, the best tag:

The road under my window is a bus-only road, that was a freight
rail track for decades (tracks ran in the center) before the buses
started to run there, then only occasionally used at night, then
disused for some years before the tracks were removed last
summer. They'll build, eventually, tram tracks where the driving
lanes are now, but then the road (emergency vehicles only) will
still be something that was a railway track.

A linear clearing with some scrubby young trees in a small but
healthy wood area nearby is also there because that freight track
was partially realigned a few decades ago (only the tracks were
removed, still railway=abandoned). If nothing is built there, the
line could, in some or several decades, become
indistinguishable; at that point it would be appropriate to change
to was:railway=rail.

Verifiability doesn't mean it's easily seen with the naked eye at
ground level, but that the next person can use any combination
of observations, previously mapped related data, and reliable
sources to make up their mind if the feature is or isn't (or
wasn't) correct. That way drawing multiple generations of past
buildings in cities with a long history (an example mentioned
here) wouldn't be verifiable, because even if some preindustrial
maps are suprisingly accurate, the sources don't have enough
accuracy to tell how their location relates to the interim
incarnations or to the present day data. As a counterexample,
though, the city has built, in a park, a square sett surface at
the location where the first town hall was up to 1640, based on
their research. It's a highway=pedestrian, with a note tag
explaining the historic significance, and was:building=townhall
and end_date:building=~1640. If old documents told of another
building with enough details to relate its location to the
townhall, there shouldn't be no harm in marking that area
likewise; it shouldn't matter if I were to pound some sticks in
the grass at the corners to justify tagging those first, and then
tagging "oh btw these also demarcate the site of ..."

I believe rail data has become the main argument in these
discussions, because the "better" maps became commonly
available (very) roughly at the same time as the first railroads
were being built and they leave much more partial traces than
any or most other features.

IMO even in a fully cleared new construction site, if on both sides
of that area there are abandoned or in any way identifiable ends
of a present or former railway, the was:railway=rail way through
the new development won't hurt anyone; the ends combined with
other sources verify it did go through the area. Only when the
whole subnetwork of rails has been completely erased, it's fair
to move the data elsewhere - before that, it relates to and tells
something about the remaining parts or of their remains.

Somebody suggested that storing historic administrative borders
would be excessive or out of line; I say if anybody can deduce
and map them, they deserve to be available and related to present
day features. E.g. municipalities do merge and sometimes swap
plots, and any research of history would benefit if the boundaries
of former municipalities were available. We just wouldn't tag them
as boundary=administrative, but was:boundary=administrative
(with relevant date tags). Even old country borders aren't available,
but historic documents may include description of their boundary
stones (some still exist, historians have documented these) or
geographic features used as the defining points. The sections
between those points can be very, very long, so if they were
commonly editable and publicly stored they would tell something
about the current data: "this house is built in a location that was
in country X  up till year 1823, in Y between 1823 and 1917 and
in country Z since that" - an attribute of the present day world
just stored in the least complex way, i.e. in a similar way as the
present day borders. The former border is still where it was.

The borders don't usually have too many turning points, and they
follow the same lines even for centuries, so these won't make
editing more difficult, in any detectable amount.


More information about the talk mailing list