> If the light is purely a warning to both traffic and pedestrians, then it's not crossing=traffic_signals. If it controls both traffic and pedestrians (control as in indicating whether they should halt or proceed) then it's crossing=traffic_signals. Your situation of warning lights for traffic strikes me as bizarre and could be argued either way. Or maybe it needs crossing=insane_signals.
I think it should be a completely separate tag that specifies exactly what traffic is being controlled, and by what, with optional levels of specificity, like most other tags. I would rather have crossing:stop_sign=no and crossing:street_signals=hawk than crossing=traffic_signals, for example, because at least everyone can know what the former means. They can even be presented as simple yes/no questions. > I think the clue is in the name: crossing=TRAFFIC_signals. Also CROSSING=traffic_signals. And crossing=traffic_SIGNALS. (...) The clue-based strategy is clearly not working. There are people on this mailing list with very strong opinions that disagree regarding the meaning of this tag. It's easy to sympathize: we've got a tag about a *crossing* specifying *traffic* signals, but not the exact kind of traffic (pedestrians are also referred to as traffic) nor signal type aside from it being lights. There is then all of this implied interpretation regarding "controls" that always comes up but is both unstated in the wiki and disagreed on by all. > Problematic. What would crossing:traffic_signals even mean other than traffic lights with an uncontrolled pedestrian crossing. It would mean there are signals directed at traffic for this crossing. This is information that cannot be reliable determined from the current schema because mappers don't know what crossing=traffic_signals means. Many mappers think it means pedestrian-specific signals. I've met them in-person! Regarding an "uncontrolled pedestrian crossing", keep in mind that the wiki calls a crossing "uncontrolled" if it just has markings on the ground, which disagrees with what uncontrolled actually means everywhere else. > Going by what we have at present, it's traffic lights with an uncontrolled crossing. Uncontrolled because nothing except common sense tells pedestrians when to cross. (...) In that image there are lighted signals telling pedestrians when to cross. It's those black boxes lower on the poles. > I don't see the markings as necessary. They're a secondary element that may or may not be present depending upon jurisdiction. Then crossing=traffic_signals does not imply a marked crossing and we're back at my previous question. You can't have both: either crossing=traffic_signals implies marked crossings or marked crossings are not necessary for crossing=traffic_signals. They are contradictory statements. If we accept that markings are not necessary for crossing=traffic_signals, then this tag has the original fundamental problem: despite *other* values describing markings (crossing=uncontrolled, crossing=unmarked), traffic_signals does not. Data consumers are left without that information. Mappers have to grapple with competing tags: do I describe traffic signals or markings, since I can only choose one? > I'm not opposed to being able to tag a crossing as having both lights and markings but I don't see it as necessary (or even very useful). Making such a decision on behalf of all pedestrians seems a bit premature. Do all pedestrians, of all needs and preferences, share that dismissal? Someone with a vision impairment might recognize a ladder crossing more easily than a 20 cm light 40 meters away. Having a place where pedestrians should be and cars should not is valuable information for a wheelchair user (and I'd say most pedestrians). The ground markings are also considered a right-of-way signal, depending on jurisdiction. They also improve pedestrian safety. As a pedestrian, *I* want to know that information. > When to halt or proceed is controlled by the lights, and those would continue to be the controlling element even if somebody maliciously removed the road markings. Unless someone removes the lights or they break down and don't get fixed. Both can be removed, I'm not sure why this is unique to the situation of crossings. "When to halt" is information primarily for planning trips that use street traffic, which is the origin of this tag: estimating when cars have to stop. > The wiki is clear: just road markings. No lights. No crossing guard. Just markings. The wiki says that uncontrolled means no traffic signals (whatever that means - clearly there's confusion) and there are markings. It doesn't actually say anything about lights regarding "uncontrolled". It also says it's equivalent to an American crosswalk. Doesn't say anything about crossing guards, because in that case it is, sensibly, a separate tag: supervised=yes/no/*. And you can see that the actual meaning of "uncontrolled" creeps in. The wiki is not actually describing an uncontrolled crossing, it's describing a marked crossing, but every time there is a discussion of crossing=uncontrolled everybody chimes in with their own ideas of what it means for a crossing to be controlled. The OSM data is then punished when mappers attempt to understand the term "uncontrolled": more transit knowledge could actually lead to worse data. > Ummm, we're into pedantic parsing again. Yes, it's a crossing. A crossing that has lights controlling both vehicles and pedestrians. To be marked as crossing=traffic_signals it must be a crossing which has lights controlling both. Lights that may or may not be mounted on the same pole but which operate in synchrony. The distinction between "is a" and "has a" is very important for this tag, it's why crossing=* is a mess. The values should be "is a", because they are implicitly orthogonal: you either tag crossing=uncontrolled or crossing=traffic_signals. The crossing "is" an uncontrolled crossing. The crossing "is" a traffic_signal crossing. But this is wrong: those two values are not truly orthogonal, they make statements about different things. A crossing *has* ground markings. A crossing *has* traffic signals. They are separate properties. I don't know why this distinction is pedantic. Seems important to me. On Sat, May 11, 2019 at 5:22 AM Paul Allen <pla16...@gmail.com> wrote: > On Sat, 11 May 2019 at 01:09, Nick Bolten <nbol...@gmail.com> wrote: > >> > I would not expect to see something like that, in any of its regional >> variations (green walking person/red stationary person in much of Europe) >> without related lights controlling traffic. >> >> So, in the case of a pedestrian warning beacon, which does not control >> traffic in the cases you've mentioned, how would you tag the crossing? >> crossing=uncontrolled? Even thought it has pedestrian-facing lights *and* >> lights intended to warn traffic about pedestrians? >> > > If the light is purely a warning to both traffic and pedestrians, then > it's not crossing=traffic_signals. > If it controls both traffic and pedestrians (control as in indicating > whether they should halt or > proceed) then it's crossing=traffic_signals. Your situation of warning > lights for traffic strikes me as > bizarre and could be argued either way. Or maybe it needs > crossing=insane_signals. > > >> >> There's another user in this thread who thinks the polar opposite and >> that the lights directed at traffic are irrelevant. Who is correct? What >> does the tag mean? >> > > I think the clue is in the name: crossing=TRAFFIC_signals. Also > CROSSING=traffic_signals. And > crossing=traffic_SIGNALS. It is a crossing for pedestrians which has > signals for the traffic. Of > course, going by the name alone you could argue there are signals for > traffic but not for pedestrians > but that is handled as traffic lights plus crossing=uncontrolled. > > >> This is making me think that my other proposal should be revised so as to >> separate out pedestrian signals from street traffic signals entirely. >> Something like crossing:pedestrian_signals=yes/no/(type) and >> crossing:traffic_signals=yes/no/(type) >> > > Problematic. What would crossing:traffic_signals even mean other than > traffic lights with an > uncontrolled pedestrian crossing. > >> >> > Not necessarily. Most countries there probably is something, if only >> tactile paving for the blind. >> >> I think I miscommunicated - I'm referring to that intersection in the >> picture only, where the crossings are marked with the ladder pattern. >> > > Going by what we have at present, it's traffic lights with an uncontrolled > crossing. Uncontrolled > because nothing except common sense tells pedestrians when to cross. > There are no lights > telling them when they can and cannot cross. There is no crossing guard. > It's just a place > where crossing is legal (in the US it may be illegal to cross where there > are no markings; in > the UK it's legal to cross without markings) and there also happen to be > traffic lights > controlling traffic alone. > > Earlier, the necessary conditions for crossing=traffic_signals were solely >> (1) signals that control pedestrians and (2) signals that control street >> traffic. But now, at this intersection, crossing=traffic_signals implies >> markings? This is a contradiction. It is simply not possible for both to be >> true. So you can see my dilemma in trying to use such a tag. >> > > I don't see the markings as necessary. They're a secondary element that > may or may not be > present depending upon jurisdiction. At most they're a warning that there > are traffic signals > controlling a crossing, a warning for people looking down at the road who > might otherwise not > notice the lights. When to halt or proceed is controlled by the lights, > and those would > continue to be the controlling element even if somebody maliciously > removed the road markings. > > Keep in mind, all I want to describe is whether a crossing has markings >> and whether it has pedestrian signals. That seems like something just about >> any person on the street should be able to answer, but the schema makes it >> difficult to tag. >> > > I'm not opposed to being able to tag a crossing as having both lights and > markings but I don't > see it as necessary (or even very useful). There are crossings that are > only markings and there > are crossings that are controlled by lights. > > >> >> > I would say that both pedestrians and traffic have to be controlled. >> Controlled pedestrians and >> uncontrolled traffic is insane. Controlled traffic and uncontrolled >> pedestrians is traffic lights. >> >> Not according to most definitions of "controlled", in terms of traffic >> lights. A stop sign is a form of traffic control. Also, someone in the >> other thread claimed that dropped curbs were a control. Someone in this >> thread says a marked crossing is a control. Nobody agrees on what a control >> is in OpenStreetMap, so how can we ever trust data for "uncontrolled"? I'd >> guess that almost nobody is using it for anything other than a delay on a >> router (car might have to stop) or some generic visualizations of feature >> density. They can't, not reliably. >> > > The wiki is clear: just road markings. No lights. No crossing guard. > Just markings. > >> >> > The pedestrian-facing lights and vehicle-facing lights don't even have >> to be on the same pole, but they should be positioned such as to control >> the pedestrians and traffic at a crossing and be operated in synchrony by >> the same controller. Together they constitute a single crossing. >> >> I wholeheartedly disagree. A crossing is not the signals. A crossing is >> where pedestrians cross the street. A crossing can *have* signalization: >> signals are a property of a crossing. This is similar to how a crossing is >> not an island and why crossing=island was a bad idea. >> > > Ummm, we're into pedantic parsing again. Yes, it's a crossing. A > crossing that has lights > controlling both vehicles and pedestrians. To be marked as > crossing=traffic_signals it must > be a crossing which has lights controlling both. Lights that may or may > not be mounted on the > same pole but which operate in synchrony. > > > Oh, and you can have two independent crossings within a few yards of >> each other which handle one direction of traffic flow on a road with >> several lanes (...) >> >> As footways, I would map this as three elements: all are >> footway=crossing, the central island is crossing:island=yes, the other two >> are... well, I don't know, really. That's what I'm asking questions about. >> Maybe crossing=traffic_signals. >> > > I refer you to > https://www.gov.uk/guidance/the-highway-code/using-the-road-159-to-203 > rule 197 and > https://www.gov.uk/guidance/the-highway-code/rules-for-pedestrians-1-to-35 > rule 28. You may find the other rules useful. Yes, these apply to > crossings in just one > country, but your new tags will have to be able handle them correctly (and > they may be > useful in other countries). > > -- > Paul > > _______________________________________________ > Tagging mailing list > Tagging@openstreetmap.org > https://lists.openstreetmap.org/listinfo/tagging >
_______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
