[Tagging] Use of highway=track vs highway=service cemeteries, parks, allotment gardens, golf courses, and recreation areas

[Minh Nguyen]

Vào lúc 03:15 2021-02-27, Bert -Araali- Van Opstal đã viết:
> there were indeed voices that preferred highway=residential, but it has 
> only disadvantages. :)
> People expect from a residential road that they can drive on it with 
> their car, also these narrow alleys are typically not the only kind of 
> roads in these centers, and it made a lot of sense to have a different 
> tag in order to see the difference on the map, avoid routers sending 
> people into these roads etc.
> 
> I understand. But as I understand they do support other vehicle 
> transport like motorbikes and bicycles, modes of transport which get 
> more and more promoted and supported by legally enforced access 
> restrictions (with traffic_signs), even on wider roads.  It looks weird 
> to me to use a highway=service tag for that purpose, that's why here we 
> look at the highway=service tag as being generally used for roads that 
> have no specific general "public" character, a tag that has service=* 
> values that mostly describe a certain use or purpose, not covered by the 
> other highway=* tags. I know, people already pointed me out that is also 
> not the correct interpretation as it was intended, but maybe that has 
> grown historically a bit out of order due to reluctance to add a new or 
> attribution to the already existing highway tags ?
> That's why we locally tend to map these as residential roads. Recently, 
> we started to use the narrow=* key together with width=* to provide data 
> for such "alleys",  We also had a group which liked to prefer to use the 
> access keys to tag not only legally enforced access but also physically 
> "enforced" access. Is that practised the same way elsewhere in the world ?

Sure, here are some situations where I'd tag access based on physical 
characteristics:

* If a golf cart path is tagged highway=service, then it needs 
motor_vehicle=no to keep street-legal vehicles from getting stuck or 
potentially mowing down pedestrians.
* A particular barrier=bollard could be motorcar=no regardless of 
signage or legal prohibitions, just based on the width on either side of 
the bollard.

> It would make it easier also for routing software in my opinion, 
> especially for motorists and cyclists, they should look at the physical 
> highway keys. Anyway, my perception is that we don't tag for a specific 
> data use or rendering.

Certainly we'd want to avoid choosing tags that favor one use case at 
the expense of another, or one specific product at the expense of 
others, but we do need to facilitate use cases such as routing.

Ultimately, users expect routers to avoid service roads, including 
alleys, except for hyperlocal use (when the origin or destination is 
directly accessible from the service road). In my experience working on 
turn-by-turn navigation software, one of the most common kinds of user 
feedback I encounter is overuse of alleys, either as a "dive-bomb" 
around traffic or because a geocoder placed the destination closer to an 
alley than the through street out front.

The prevalence and prominence of an alley differs a lot even within a 
country. But in most places in the U.S., even if a house or business 
backs up to an alley, it would be inappropriate for a visitor to show up 
at the back door. (Mapbox's old headquarters in Washington, D.C., was a 
curious exception with "Rear Entrance" even in the address.) In many 
cases, a human can easily spot an alley just by paying attention to 
patterns in the street grid.

Besides a suboptimal user experience, overuse of alleys can be a safety 
issue. In the U.S., alleys aren't allowed to have the same safety 
features as through streets -- no stop signs where two alleys intersect 
or warning signs about children at play -- because the signs would be 
legally unenforceable. [1]

In theory, it would be possible to craft a routing profile that 
penalizes narrow or substandard roads for some local definition of 
narrow or substandard, but there are practical, algorithmic limits to 
how much context a router could incorporate into these decisions and 
still remain performant enough for real-world use. To a human, a 
one-lane, unstriped road within a city may be obviously an alley, but a 
road with identical characteristics just outside the city limits might 
be a through street.

> We have these also, and I think you find them across the world mostly in 
> historical cities. But how do you distinguish the alleys which are 
> accessible for the general public, and not clearly signposted as such, 
> and the ones that have a private use only because they give access to 
> just a single house. F.i. we have many alleys in towns but they are not 
> intended for public use, they pass at the back side of building rows. 
> They are not intended for "public" use but to allow people to access 
> their homes from behind.  SO we map them as highway=service.  People do 
> use them as passages, especially pedestrians, because the main roads get 
> to congested and become dangerous because of misbehaviour of motorists, 
> like driving on the footways.  We look at it though as non-legal 
> behaviour which we don't tag, so we don't tag them as residential roads 
> but as service roads.  They would however perfectly classify as alleys. 
> As long as someone doesn't put a barrier or a clear signpost that a road 
> is private, we consider all roads as publicly accessible. No publicly 
> accessible highway has a specific highway=service tag.

This sounds to me like motor_vehicle=destination, but I'd contend that 
users already expect highway=service service=alley to imply 
motor_vehicle=destination anyways.

[1] http://lancastereaglegazette.newspapers.com/clip/72275634/ 
http://lancastereaglegazette.newspapers.com/clip/72275654/

-- 
minh at nguyen.cincinnati.oh.us