> Which seems to be precisely the opposite of how most people interpret it.
Which is very bad, because those people are all diametrically opposed to the wiki definition that, for all its problems, been around for about a decade. To me, this says that there is likely a lot of bad data out there. I'm going to skip the remaining, "markings don't change the interaction" claims given they've already been addressed. > Worth mapping for the benefit of the visually impaired, but not by redefining current usage. It can't be mapped under current usage. "Worth mapping" implies fixing this tag. > So then we need marked_and_not_controlled_by_lights and marked_but_controlled_by_lights. Which is fine, as long as you don't redefine current usage, because that would cause major problems if it went through (which it almost certainly would not). Those tag values would be redundant with and surely deprecate traffic_signals. > Explain how your proposal would significantly reduce errors It's in the wiki under the proposal. I also walked through it twice in my response. > Aerial mapping a new crossing with stripes is going to result in a marked crossing either way. Nope. If it has pedestrian signals, which marked crossings clearly can, the current schema calls for it to be described with crossing=traffic_signals, which erases marking information. Look at imagery outside of the UK or some of the examples posted in other threads. And if someone maps crossing=uncontrolled from aerial imagery of a crossing based on ground markings, the crossing tag is set, so that crossing won't be "fixed" through usual QA or quests based on missing data. > Currently if I edit an existing crossing because I see stripes in aerial imagery I see from the tag list that it's already been marked as having lights and would have to change that to being uncontrolled, so I do further checks (has the type of crossing changed, can light-controlled crossings have stripes in this jurisdiction). The example I've given is the case where a mapper is creating a new crossing: they're armchair mapping and can identify the markings, but signals are hard to gauge from an aerial view. In this case, the crossing has already been mapped. If they are truly using the schema I've proposed, they've set crossing:signals=yes. If they set crossing=marked as well, it doesn't erase the crossing:signals tag and data consumers have sufficient data to present appropriate information to any user. If we're still using crossing=traffic_signals for some reason, then that would be an argument for using crossing:markings: don't erase information. > Wikipedia? > Like https://en.wikipedia.org/wiki/Pedestrian_crossing It makes a distinction between signalized and unsignalized crossings. This doesn't disagree with anything I said. > The tag values are unclear and misleading without referring to the documentation, which is the case of many tags, but this is how most people interpret matters (...) Hard disagree. If we went solely by the various responses in these threads, I would say there is no majority opinion on their use. Example: the definitions recently endorsed by Thorsten and yourself explicitly disagree with the wiki because they keep involving legal right of way. > That is the current usage, and the current implication of the wiki; redefining it would cause big problems. Let's look at what redefinition would be: taking tags created under one impression of the tag's meaning and then everyone else wholesale declaring that it now means something else. This is actually what is implied by the wide variety of definitions we've seen in these discussions. People are redefining each others' tags all over the place, reinterpreting what "uncontrolled" means entirely separately from the wiki. My proposals do the opposite: they create new tags with clearer meanings and don't touch the old data. Addressing the mass of "bad schema" data is left to regionally-specific efforts that could range from tasked reviews (probably a good idea in many places, given the diversity of interpretations of the current schema) to machine edits (something I could certainly justify for my area). > Suppose we wanted to replace landuse=grass with landcover=grass. (...) It would require a mass edit (...) It wouldn't. It could be used in some cases, but is not necessary. It's not even appropriate if the tag being fixed was ambiguous because that calls for manual review. There are many other tools at our disposal beyond machine edits. MapRoulette, Street complete, a variety of QA tools. I could create an OSM tasking manager instance to split up work in a particular area. Data consumers that can make sense of the current schema for pedestrians (haven't seen one yet) can keep interpreting the old data as best they can. They can support both during a transition, because there actually isn't any redefinition. > On 2.5 million POIs, that's not going to happen. Yes, it is. That number is less impressive than you might think. Our project and direct collaborations have mapped 5-10% of crossings and represents the efforts of a relatively small number of people. There is tons of enthusiasm to tag this data that is currently going to waste. Right now, I can't give the people coming to us a tagging schema that will work in the long run. Harnessing that enthusiasm just needs tooling and an actually good schema. > >
_______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
