[Talk-us] Potential Mechanical Edit to remove access=private from Amazon Logistics driveways in NH

[Minh Nguyen]

Vào lúc 06:17 2020-09-01, Matthew Woehlke đã viết:
> On a related note: I use service=driveway (for lack of anything better) 
> for access ways to parking lots that don't have parking spaces (hence, 
> not service=parking_aisle). These are likely *not* public right-of-ways 
> (the lots themselves are usually "private"), but they are also certainly 
> not access=private.

The access roads that form the backbone of a parking lot layout 
shouldn't be tagged service=driveway. [1] You're right that there isn't 
a more appropriate, well-established service=* tag, since no one seems 
to know of a concise, intuitive term for these access roads. But there's 
also no requirement that every highway=service way have a service=* tag. 
(Shared driveways also commonly lack service=*, but there are at least 
some tagging ideas floating about.)

Routers that understand service=driveway treat it as the very bottommost 
rung in a hierarchy of road classifications, applying penalties similar 
to the penalties for parking aisles. These penalties generally ensure 
you'd never use a driveway as a shortcut between two non-driveways. But 
around a parking lot, they may lead to suboptimal routing to a loading 
dock or drop-off area, traversing ordinary parking aisles instead of 
these access roads that are likely to be wider and more traversible.

If we do someday come up with a good tag for the access roads you're 
describing, it'll be easier to identify and retag them if they lack a 
service=* tag than if they're tagged service=driveway. After all, 
service=driveway would be more appropriate for a drop-off area coming 
off a parking lot. In the meantime, it's rather nice that some renderers 
and editors give service roads deemphasize service=parking_aisle ways 
while leaving these bare highway=service ways a little more prominent by 
comparison.

> So, no, service=driveway should *not* imply 
> access=private. If anything, lacking other information, it should imply 
> access=yes just like it does on any other way, and I suspect routing 
> engines route accordingly.

This is my understanding as well. It's OK that a router generally treats 
driveways as accessible: most driveways are dead-ends, so a router would 
only lead a delivery driver down a driveway if they have a package to 
deliver to that house. (Clearly Amazon's intention is to know how to go 
down the driveway instead of stopping at the curbside mailbox.) On the 
other hand, if the driveway isn't a dead end, a penalty should prevent 
it from being used as a Waze-style shortcut around a more easily 
traversible public road.

> This, BTW, is a large part of why we're having this conversation in the 
> first place. The problem with overusing access=private is that we're 
> effectively teaching routing engines to ignore that, which makes such 
> tagging much less useful.

Exactly. As a mapper, I'd be concerned about a logistics company such as 
Amazon needing to route to the front door, so to speak, and therefore 
tossing out a convenient way to distinguish between two common kinds of 
driveways that can't be used in the same manner.

Vào lúc 12:56 2020-08-31, Kevin Broderick đã viết:
 > Second, I'd like to point out that there *are* driveways in New England
 > that are actually public right-of-ways.
 > https://www.openstreetmap.org/way/19685143
 > <https://www.openstreetmap.org/way/19685143> is one such example; the
 > southernmost portion of the way is arguably service=driveway, except
 > that it is actually a public right-of-way that continues south,
 > eventually connecting to Lincoln Gap Road. While they are certainly the
 > exception and not the rule, the number of such setups in Vermont is
 > non-trivial due to the ancient roads laws there. There are probably some
 > similar cases in New Hampshire and possibly Maine, I believe, but I
 > can't cite any off the top of my head (the documentation of unmaintained
 > public-right-of-ways isn't as good as it is in Vermont, making things a
 > bit more murky).

Penalties aren't absolute prohibitions, so if there's no other way to 
get to that public road, or if the alternative would take much, much 
longer, the router will hold its nose and use the driveway. But it can't 
do that if the road is marked as impassable via access=no or restricted 
via access=private.

It can be a bit unintuitive to think about routing penalties when 
mapping. During a navigation mapping workshop at State of the Map 2018 
in Milan, my colleague Kajari and I walked through a very basic 
application of a routing penalty to illustrate the concept. The workshop 
wasn't recorded, but the notes and slides are available at [2], with the 
explanation starting on slide 12.

[1] <https://wiki.openstreetmap.org/wiki/Tag:service%3Ddriveway>
[2] 
<https://commons.wikimedia.org/wiki/File:State_of_the_Map_2018_-_Navigation_Mapping_Workshop_(with_notes).pdf?page=12>

-- 
minh at nguyen.cincinnati.oh.us