One is that “all traces are gone” is in the eye of the beholder. A
subject specialist may be able to see traces where a casual remote
mapper couldn’t. In this case we should always defer to the subject
specialist. OSM thrives (I’m repeating myself here) by collecting the
knowledge and enthusiasms of subject specialists.
I disagree. If you add stuff to OSM that only subject specialists can
understand and handle, it will get broken, because like it or not, OSM
is a project of mediocracy. A project for the masses. Everyone can and
will edit it, and creating things in OSM that carry warnings “DO NOT
EDIT! SUBJECT SPECIALISTS ONLY!” curtails the ability of the many to
participate. If there’s stuff in OSM that I do not understand (and
cannot verify on the ground because special knowledge or equipment is
required) then I will either refrain from touching it (even if there
would be good reason), or I will accidentally break it.
That’s also why we try to have human-readable tags - we must build our
project so that everyone can participate, and not have “here be dragons”
areas that only specialists may enter. It’s bad enough already with
things like public transport or complex junctions with turn lanes and so
on. But adding stuff to OSM where any average person going to the place
would say “I see a few trees and a cycleway” but instead it’s a former
railway which anyone who can smell the amount of iron dust in the ground
will attest to - that won’t work, it will break, and it must break if
non-iron-sniffers are to be allowed to edit.
I get the idea in principle. But I have never found a lone pixel-width railway=dismantled to encumber my editing. Not like administrative boundaries which are an enormous PITA when trying to do any sort of highway or waterway editing. Whoops, this river meander has been cut off due to floods, I’m now going to have to spend half an hour repairing the various boundary relations after remapping the river.
I’m sure you wouldn’t have posted if you didn’t find it a problem but… it’s so far from my experience I’m just surprised. Hey ho.
certainly “having to install JOSM and fiddle with its filters” is not the path to
“harnessing the many different enthusiasms of millions of contributors”.
I am going to frame this and put this up on the wall as “the moment when Frederik accepted that JOSM is not necessary for mapping”
Remember, the licensing barrier only goes in one direction. There’s nothing preventing the two datasets from intermingling in postprocessing or even while editing OSM. At a glance, this sounds like the sort of conflation task that many data consumers already perform when joining OSM data to other data sources. For example, for better or worse, Overture Maps takes an OSM building, falling back to a non-OSM building if it doesn’t intersect any OSM building. Overture doesn’t need to ram the entire Microsoft building dataset into OSM or vice versa in order to obtain a more comprehensive building layer. Microsoft buildings are ODbL-licensed, like OSM, but other maps use the same approach with public domain address and waterway datasets as well.
Granted, this conflation strategy isn’t foolproof. Sometimes OSM mappers have gotten the scoop on a demolished building before other datasets have updated, so in the absence of anything in OSM, Overture revivifies the building. This is why I’ve tended to retag demolished buildings as demolished:building=*, in the hope that a data consumer will consider it in its conflation strategy. If this is anyone’s reason for mapping completely obliterated railways – that a data consumer would otherwise backfill it from a third-party layer of severely hallucinated ML-detected railways based on outdated imagery – then I concede the point. But then we get into the question of how long to keep a demolished building around before it blocks a newly built building that we didn’t notice.
If problem would be limited to this then it would not be significant trouble…
But: sometimes these are far more dense networks or long gone railway buildings are also being mapped.
Or entire railway yards.
Or this likely not surveyed on the ground edit adding entire network of logging railways. Likely at least in part having no trace anymore on the ground.
Also, what worries me is if that is being seemingly accepted and not removed in time then we will have more classes of similar data added by people thinking it is welcomed.
As I understand OHM is fine with mapping 40 different reconstructions how specific castle could look like in XV century. But one of nicer parts of OSM is that in the end if there is conflict then on the ground situation resolves it.
That is impossible if people argue which reconstruction is more accurate (not sure how OHM deals with it).
And I live in a city with more than thousands year of history, with so many destroyed buildings that land was raised by meters and to go into older churches after centuries of buildings being torn down you go downstairs rather upstairs.
Mapping all buildings ever existing - or just road network changes over hundreds years - would ensure that data editing would be impossible. And even just mapping now gone cargo railway stations would protect some areas from editing by anyone who is not a JOSM expert.
In the end we would remove it - but we will have even more disappointed people that would get months/years of their effort removed from OSM.
Did you try asking the contributor there (a) what is left on the ground (b) if not, whether they considered OHM and (c) if they did why they didn’t use it?
Yes, changeset comment 10 days ago, their latest edit is 22 days ago so for now I am waiting before doing anything more like reverting this edit. That is my typical first step in such cases.
If they will reply in way indicating no survey being made (mapping from old maps alone) or give reply that these railway are utterly gone I will surely recommend OHM.
Sure. I’m not worried about "might happen"s (I think you and I have had this exact conversation on the mailing lists in the past ), nor am I interested in justifying widespread mapping of railway yards where no traces remain.
Essentially the proposal from the Lovely Users Of A Particular Chat Client™ is that railway=dismantled and railway=razed should be entirely deprecated (“these two tags should not be in the database”). I have said that there are clear circumstances in which these tags are useful, and that a blanket deletionist argument risks driving away more contributors than any benefit to the map. That’s it.
I can try doing this, but in typical case I have no idea at all what is the source and whether this railway actually ever existed. Or mapper replied that they plan to do it in OHM themselves.
(also, not sure whether I need to get mapper permission before import if they traced old map - another question is about tracing using also OSM data or Bing imagery which add additional unfunny legal complications)
So far I had no case where I would have opportunity to do this but I am open to doing this.
nobody is proposing to map long gone buildings with no traces left, on the other hand, also living in a city with a long history, a lot of historic buildings are actually still there, somehow, as basements of “current” (typically also historic) buildings. Sometimes filled with earth and rubbish and then rediscovered and excavated. As long as they are not accessible, we do not map these remains usually (as we don’t know about them, cannot survey them, although they are expected to be found in the whole inner city), but in a few cases these are open for public visits, and can naturally be mapped (although it is not typically done in greater detail either, because underground structures are generally harder to map, and it would basically amount to indoor mapping).
The main question of contempt seems to be about how evident these structures must be. Is it sufficient that a specialist can recognize them, or do we only want stuff that my 9 year old can identify, ideally with an information board (purists have argued even with the information board, only the board should be mapped). I’m with the specialist fraction here. While I see Frederik’s point that a map with millions of contributors likely will have to accept some tendency for mediocracy, I do not see how we cannot make contributors accept and tolerate specialist data (by type and depth) along their own contributions - live and let live. Ultimately all our data is “specialist data” (or has a potential to become it), even editing the highway network (for example the subtle differences in bicycle tagging according to the country and signage, but also a lot of other details) requires more interest in the details of traffic law than I assume most drivers have kept a few years after passing the test and getting their permit (if they have one at all). A former railway that may be hard to spot for the untrained eye will not impede someone to fix a name or a housenumber, or add a new shop or building, it may at most make them raise an eyebrow, and who knows, maybe on their next survey will discover some interesting traces they otherwise would not have noticed.
Isn’t there a middle ground here? Imagine two different scenarios:
A subject specialist recognizes an embankment of a certain shape is clearly an old railbed. A non-specialist mapper can see the embankment but doesn’t recognise it as an old railbed. However when the specialist gives them a short explanation the non-specialist can see how this makes sense. After all there is a terrain feature there that they can see.
A subject specialist recognizes an old railbed one side of a town and old rail bridge abuttments on the other side. Based on the alignment of these and the orientation of buildings and streets in between, they are able to extrapolate the likely path the railway used to take through the town even though now many newer buildings sit directly where it used to be. Upon hearing this explanation, the non specialist says “I can see the embankment and the abuttments, but in between all I see is buildings and streets. I do not see the faint hints you are relying on”.
To me, scenario 1 is 100% reasonable to include in OSM because although a non-specialist may not initially know what these features are, they can see them and a specialist can easily provide an explanation. I see scenario 2 as much less reasonable to include in OSM (the part through the town) because it requires far more special knowledge and a trained eye for small details that cannot be so easily explained to the non-specialist. OSM absolutely does rely on subject specialists, but also generalists, and the generalists should be able to at least understand the evidence presented by the specialists.
Not to pick a fight by asking, but why is mapping in OSM the USBRS a “sin” of any sort? True, this bicycle network is only “lightly signed,” (estimates are “about 25% of routes are signed, and growing”) but it is a distinctly “real” transportation network, designated by state Departments of Transportation and cataloged by AASHTO (the USA’s authority for national transportation network numbering protocols). No “sin,” no harm, no foul. And its processes for emergence of new routes has been painstakingly documented in our wiki for over a decade, with apparently wide approval (except for me reading just now that Richard might consider this a sin). Brother, please help this fellow mapper understand what is sinful about the USBRS!
(Back on topic): I’ve felt for a long time, and maybe have articulated clearly this, maybe haven’t, so I will now. It would be ideal if all railway=* [abandoned, dismantled, razed] (but not disused) were to eventually migrate to OHM, where it could be rightly said “these data belong.” I realize that isn’t going to happen (neither overnight nor wholesale) without a coordinated effort. That effort could be organized in OSM as a planet-wide / railway-mapper-enthusiast project, but I don’t recommend anybody hold their breath for that to happen. It is, one among many, a “quirk of OSM” (and our history of rail mapping) that we do things this way. Is it messy and inconsistent? Yes. Are there better ways to do things? Yes. Will we? Well, time will tell. Our map is never really “done,” is it? It’s usually pretty darn good, and it always seems to have room for improvements. I’m OK with that. And, “onward with improvements.”
And maybe we can use that time to discuss another burning issue: what to do about old airways that can be observed through remaining aerial lighthouses?
Because it requires additional sources to make sense of what is observable on the ground? Same for hiking routes, and I’ve wondered a couple of times about why and how the community accepts to slightly distort the observability principle.
It seems that the “amount of sin” is the complexity of the process to gather external sources and make deductions from ground observations to statements expressed as tags. And that tolerance is probably related to the “amount of sin” divided by the number of people who make use of the result.
Mmm, no. Simply “follow the map” and you’re good to go. No additional or external sources required: that is the entire point of having these route data in OSM. Totally “self-contained.”
What I’m saying is that our relations are a construction based on logical deductions on the ground plus additional data from operators when there is some doubt. For most hiking routes I’ve worked on there has been a time when an external source was needed to create a continuous line in OSM, and my recent work on EuroVelo suggests it’s the same for cycling.
I stand by my theory that the amount of deduction and external sources (historical sources in the case of old railways) needs some compensation to be accepted by the community.
Out of curiosity, when an Ordnance Survey map labels a “Dismantled Railway”, does that mean there are obvious traces on the ground, or that the locals routinely navigate based on where the railway used to be, or that an older OS edition depicted the feature in its active state? If someone is attempting to replicate the level of detail in OS maps in OSM alone, can OSM’s coverage of (1) suffice, or do we need (2) as well?
Things are a bit different in USGS topos: they don’t depict dismantled railways per se, so when I map these lines in OHM, I generally have to pull up older quads from the 1960s or ’70s when they were still active. On the other hand, the topos accrue lots of former boundary and survey lines. I’m sure they serve a purpose, especially out west where boundaries aren’t as obvious on the ground but still affect access rights. I’ve relied very heavily on these same 20th-century topos for mapping the vast ranchos (Spanish and Mexican land grants) that covered my area during the 18th and 19th centuries.
These ranchos have lent their names to a great many things in my area, including parks, shopping centers, and the local system of bus routes.[1] They’re easy to map from current topos, and with enough context from research at the desk, you can even map them based on present-day roads and streams. However, we’ve refrained from mapping them in OSM as parcels per se. We’ve mapped the few remaining farmhouses, but the only way you’d know about their origins is by following the link to Wikidata, and from there to OHM. So far, no one has complained, but then again, there aren’t dedicated local rancho historians like there are for railway history.
San Jose Mercury News, 24 November 1974, p. 394. ↩︎
I wouldn’t say I’m a good one, but I’m an amateur “rancho historian” around here (California) too, agreeing with you that ranchos and Mexican land grants color (vividly!) many things named (on maps and older maps) around here.
), nor am I interested in justifying widespread mapping of railway yards where no traces remain.
