[Tagging] RFC: pedestrian crossing as an area

[Minh Nguyen]

Vào lúc 09:01 2022-01-13, Brian M. Sperlongano đã viết:
> My gut reaction is that pedestrian crossings can be adequately modeled 
> as a center line and width (similar to roads with width and/or lane 
> tagging).  While mappers are free to innovate and come up with new ways 
> to map in increasingly greater detail and specificity, I think that the 
> "approved" tag that ends up on the wiki after an approved proposal sends 
> a specific signal to mappers.  And that signal is not just "this is the 
> community recommendation on how to tag X" but also, and the point I want 
> to express here, that "feature X should be mapped".
> 
> I feel that a pedestrian crossing areas (in US English I'd describe this 
> as "the area of a crosswalk") fits into a category that's increasingly 
> described as "nano mapping".  I seem to recall somebody proposed to tag 
> the area of individual road markings - the dashes and arrows and so 
> forth painted on the road, and I see this as a similar "nano" feature.  
> I'm confident that there are plenty of other things left to map in the 
> world before we need people individually tracing painted road markings 
> just to keep busy, so there should be no rush to apply that green 
> "Approved" label.

In practice, the "Approved" status doesn't dictate anyone's personal 
priority in mapping or consuming a tag. It's sort of a protection 
against a mapper getting bogged down by arcane tag design questions when 
all they want to do is map something they see. It would be a different 
story if, say, an editor were to color-code all the tags in the UI by 
their approval status.

If we want the status to carry more weight on a personal level, we'd 
have to conduct a review of the wiki's historical approvals to achieve 
some sort of editorial balance. There are a *lot* of cultural and 
lingustic assumptions about priorities baked into in our tagging system, 
especially if you filter it by nonorganic criteria like approval status. 
But an audit of approved tags would be a big project and for 
questionable benefit, considering that there are other ways to influence 
mapper behavior, and data consumers have their own motivations beside 
maximally hewing to the wiki.

Regarding the proposal at hand, I don't think pedestrian crossing areas 
are as blue-sky as they might've seemed three or four years ago. When 
this topic came up on this list last May, I gave some examples of 
crosswalks in parking lots of big box stores, which have to be 
represented as multiple crossing nodes for topological reasons. [1] It 
would be nice to visually collect them into a single geometry, not 
unlike how we visually associate the various components of a bridge 
using a man_made=bridge area.

Enough pedestrian infrastructure has been mapped that it's only natural 
for some mappers to want to go into more detail, especially for edge 
cases that aren't as straightforward as a buffered centerline. 
Proprietary road datasets in some regions like Japan have long featured 
this kind of detail, and it's starting to become more common among 
global datasets too.

Our standards as a project have been trending towards more detail for 
years. For a sense of perspective, the legend on osm.org still says 
building=* is only for "significant buildings" [2], a line unchanged 
since 2009 [3], but tell that to anyone who's ever participated in a 
building import. It wasn't too long ago that private driveways were only 
mapped in exceptional cases, and U.S. mappers were actively engaged in 
deleting driveways that had been accidentally imported from TIGER, but 
then Amazon Logistics came along.

None of this means a novice mapper would be hounded for not mapping a 
crosswalk as an area, even if we approve a tag for it. But it would 
proactively ensure that, when someone does get carried away mapping 
crossing areas, they're less likely to make a shortsighted decision in 
how to tag them.

It's interesting that you bring up road markings. Several years ago, 
someone added thousands of solid stopline markings to my city. It was 
frustrating, but mostly because they used a poor tag on many of them, 
creating work for the rest of us. (To their credit, they did follow 
through with a mass retagging.) The road_marking=* key may be 
underdocumented and "highly experimental", but its relatively 
straightforward documentation is better than nothing. [4]

[1] https://lists.openstreetmap.org/pipermail/tagging/2021-May/061481.html
[2] 
https://github.com/openstreetmap/openstreetmap-website/blob/c34ed1e3706020c59aa8dc70ece41c36a8a30930/config/locales/en.yml#L2193
[3] 
https://github.com/openstreetmap/openstreetmap-website/commit/079b9fed845b429436c5e84c221154b64c500103
[4] https://wiki.openstreetmap.org/wiki/Key:road_marking

-- 
minh at nguyen.cincinnati.oh.us