Sidewalk symmetry

classic Classic list List threaded Threaded
11 messages Options
Reply | Threaded
Open this post in threaded view
|

Sidewalk symmetry

Jmapb
The wiki contains some suggestions/guidelines about when to map
sidewalks as separate footways versus when to encode them as tags on the
main road. The basic recommendation seems to be that if there's a
barrier or even a strip of grass between the two, a separate way is fine
and even sometimes preferred (and the road should be tagged
sidewalk=separate to indicate that it's been mapped as a separate way),
but if the sidewalk is directly adjacent to the road, better to just
imply it with a sidewalk=left/right/both tag.

My gut tells me that a corollary should be: If the sidewalk on one side
of the road is mapped as a separate way, then the sidewalk on the other
side (if there is one) should also be a separate way, even if it's
directly adjacent to the road due to asymmetry of sidewalk design. Does
this sound right? I certainly don't see any clean way to tag a road to
indicate that one of its sidewalks has been mapped as its own way and
the other hasn't.

(My personal feeling is that that it's better to avoid mapping sidewalks
as separate ways unless there's a compelling reason that would outweigh
the additional data clutter and routing complications. In some
circumstances -- those where walking on the sidewalk, or on a particular
side of a road with two sidewalks, has noticeably different routing
implications -- it seems like a good idea.)

Thanks for you thoughts, jmb


_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Andrew Harvey-3
On 18 April 2018 at 06:30, Jmapb <[hidden email]> wrote:
(My personal feeling is that that it's better to avoid mapping sidewalks as separate ways unless there's a compelling reason that would outweigh the additional data clutter and routing complications. In some circumstances -- those where walking on the sidewalk, or on a particular side of a road with two sidewalks, has noticeably different routing implications -- it seems like a good idea.)

It means less tags on the road which makes it easier to edit manually. A road already has:

highway, surface, maxspeed, maxweight, maxheight, width, oneway, access, lanes, turn:lanes, lit, parking:lane:left, parking:lane:right, parking:condition:left, parking:condition:right,parking:lane:left:type, parking:lane:right:type, etc.

A sidewalk also has it's own:

maxspeed (some places where bicycles can use the sidewalk and sections have signposted speed), maxweight, width, access, surface

On top of that the highway needs to be split every time just one of those tags changes meaning you end up with many short segment ways.

I realise editors can and do abstract some of this, but if we can put all those sidewalk attributes on their own ways it makes it much easier to map by reducing the complexity of the highway centerline.

It means we can use say the exact same tags on the separate sidewalk rather than prefixing them with sidewalk:left:width, sidewalk:left:bicycle, etc.

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Sidewalk symmetry

Tordanik
On 18.04.2018 05:03, Andrew Harvey wrote:
> highway, surface, maxspeed, maxweight, maxheight, width, oneway, access,
> lanes, turn:lanes, lit, parking:lane:left, parking:lane:right,
> parking:condition:left, parking:condition:right,parking:lane:left:type,
> parking:lane:right:type, etc.

The percentage of roads tagged with all these details is vanishingly
small, and will likely remain so for at least another decade.

At the level of detail that's realistically achievable in the medium
term, sidewalk tags make a lot of sense: They're easy to use for the
common case (where sometimes not even the existence of sidewalks is
mapped yet), and still allow for micromapping in pockets of unusually
high data quality.

> I realise editors can and do abstract some of this, but if we can put
> all those sidewalk attributes on their own ways it makes it much easier
> to map by reducing the complexity of the highway centerline.

Comparing the mapping styles solely based on ease of mapping would only
make sense if separate ways were able to express the same information
contained in sidewalk tags.

That's not the case, though: With separate sidewalk ways, it's
impossible (in the general case) to figure out which road section that
sidewalk way belongs to.

Not having this basic information available makes separately mapped
sidewalks unusable for entire categories of applications – sometimes
leading to worse outcomes than not having the sidewalk mapped at all.
And while you could fix that issue with relations, this would clearly
not be easier for mappers than using sidewalk tags is.

As for the original question: sidewalk=separate seems like an attempt to
solve the aforementioned issue, but it does not actually achieve this
goal – it only tells you that *some* sidewalk way belongs to this
section of road, but does not help you to find out *which* sidewalk way
that is. As such, it's not a very useful tag, and not a compelling
reason to map asymmetric real-world situations in a symmetric way.

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Mateusz Konieczny-2
On Mon, 23 Apr 2018 19:25:12 +0200
Tobias Knerr <[hidden email]> wrote:

> Comparing the mapping styles solely based on ease of mapping would
> only make sense if separate ways were able to express the same
> information contained in sidewalk tags.

Note that some information may not be expressed (or extremely hard)
with sidewalk tags.

For added fun, some people map sidewalks in even greater detail, using
area:highway (and sometimes forget to add either sidewalk tag or
sidewalk mapped as line, what is also causing problems).

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Clifford Snow
As someone who as mapped sidewalks both as metadata to an existing road and as separate ways, my recommendation is to map as separate ways. Let me explain why I recommend separate ways over the metadata approach.

Communities are starting to put emphasis on alternatives like public transportation, cycling and walking to driving for a number of reasons, For example traffic congestion. green house gases, and health benefits to name three. There is also a segment of the population with limited mobility issues that need to commute via wheelchairs and public transportation. All of this requires some form of routing. Routing via metadata breaks down at the intersection, especially complex ones. There is the routing problem that Mateusz Konieczny mentioned of what to call the sidewalk, but research scientists believe they can use the spatial proximity of the road to give instructions such as "walk on the left side of Main Street towards..." 

Separate ways also have the advantage of being able to capture data on the physical aspects of the way, such as surface material, width, smoothness, tactile pads, kerb cuts, etc. 

I've talked to some GIS folks about their sidewalk data. Some have some beautiful polygon sidewalk data. But none of the data is any good if we can't route. It's only purpose is to serve as a inventory of sidewalks.

The one advantage to the metadata approach is speed. It's much quicker to add sidewalk=left/right/both/none. But like the cities standalone sidewalk data, it doesn't route all that well. 

Jmapb - if your goal is just to map sidewalks as inventory - then use the metadata approach. I can attest that it is easier. But if you want someone to use the data, then map it as separate ways. 

There is a good website that explains the separate way approach opensidewalks.com. I know the people who put it together and they convinced me it's the better approach.

Clifford




On Mon, Apr 23, 2018 at 1:26 PM, Mateusz Konieczny <[hidden email]> wrote:
On Mon, 23 Apr 2018 19:25:12 +0200
Tobias Knerr <[hidden email]> wrote:

> Comparing the mapping styles solely based on ease of mapping would
> only make sense if separate ways were able to express the same
> information contained in sidewalk tags.

Note that some information may not be expressed (or extremely hard)
with sidewalk tags.

For added fun, some people map sidewalks in even greater detail, using
area:highway (and sometimes forget to add either sidewalk tag or
sidewalk mapped as line, what is also causing problems).

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk



--
@osm_seattle
OpenStreetMap: Maps with a human touch

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Ed Loach-2
Clifford wrote:

> There is a good website that explains the separate way approach http://opensidewalks.com
> I know the people who put it together and they convinced me it's the better approach.

I would say separate ways make more sense in urban USA where you can't cross the road just anywhere, see e.g.
http://news.bbc.co.uk/1/hi/6251431.stm

In the UK I would use sidewalk tags where the pavement (sidewalk) is only separated from the road by a kerb, and separate ways where there is something more (such as grass verge or fence or whatever). In the cases of verges I would then make sure private driveways, etc that cross the footpath are mapped so pedestrians can see the obvious places to cross without getting their shows wet should the grass be wet. You can still do things like sidewalk:surface if you want, and it appears many do:
https://taginfo.openstreetmap.org/search?q=sidewalk
Otherwise you start needing relations to show where separate sidewalk and road ways allow you to cross, or put arbitrary joining ways at intervals. Admittedly this method of mapping doesn't cope with the situation where there is a verge so narrow you can step across it without stepping on the grass.

Ed



_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Tordanik
In reply to this post by Clifford Snow
On 24.04.2018 02:17, Clifford Snow wrote:
> But if you want
> someone to use the data, then map it as separate ways. 

That's not the case, and it's a bit frustrating to read this just after
I wrote a mail explaining this point. To reiterate:

* With separate ways, we don't know which road section a sidewalk
belongs to.
* This knowledge is necessary for many applications.

For such a fundamental property, "research scientists believe they can
use the spatial proximity" is not good enough imo. It has to be
practical to obtain this relationship from OSM data.

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Lauri Kytömaa-2
In reply to this post by Ed Loach-2
Ed Loach wrote:

>where there is a verge so narrow you can step across it without stepping on the grass.

Unless you're with a walker, a pram or a stroller, or in a wheelchair.

> or put arbitrary joining ways at intervals.
Only useful where there's a real connection anyway, i.e. a route
starts from or crosses the highway; be it a driveway (garages count),
crossing, footway (or similar) leading away from road, or the intended
connection between sidewalk segments across the intersecting road. I
don't think any of these are any more arbitrary than the fact that in
intersection the two crossing ways both describe the area inside the
intersection, i.e. if the ways were expanded into areas with their
width, the areas overlap where the ways cross, but that's just the way
the model is.


--
alv

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Marc Gemis
In reply to this post by Tordanik
I wonder why those arguments always pop up when we talk about separate
sidewalks and not when we talk about separate cycleways.
AFAIK it is common practice to map cycleways as separate ways in OSM
as soon as there is a kerb.

Don't we encounter the same problems in data processing for cycleways ?

m.

On Tue, Apr 24, 2018 at 8:56 PM, Tobias Knerr <[hidden email]> wrote:

> On 24.04.2018 02:17, Clifford Snow wrote:
>> But if you want
>> someone to use the data, then map it as separate ways.
>
> That's not the case, and it's a bit frustrating to read this just after
> I wrote a mail explaining this point. To reiterate:
>
> * With separate ways, we don't know which road section a sidewalk
> belongs to.
> * This knowledge is necessary for many applications.
>
> For such a fundamental property, "research scientists believe they can
> use the spatial proximity" is not good enough imo. It has to be
> practical to obtain this relationship from OSM data.
>
> _______________________________________________
> talk mailing list
> [hidden email]
> https://lists.openstreetmap.org/listinfo/talk

_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

Mateusz Konieczny-3
25. Apr 2018 07:01 by [hidden email]:

I wonder why those arguments always pop up when we talk about separate
sidewalks and not when we talk about separate cycleways.
AFAIK it is common practice to map cycleways as separate ways in OSM
as soon as there is a kerb.

Don't we encounter the same problems in data processing for cycleways ?


What is preferable depends on what kind of processing and results one wants and also on mapping priorities.


I guess that people mapping and processing cycleways do it in way that tends to prefer mapping as a separate way unlike footways where some want sidewalk tag, some want footway lines.


_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk
Reply | Threaded
Open this post in threaded view
|

Re: Sidewalk symmetry

dieterdreist
In reply to this post by Tordanik


sent from a phone

> On 24. Apr 2018, at 20:56, Tobias Knerr <[hidden email]> wrote:
>
> * With separate ways, we don't know which road section a sidewalk
> belongs to.
> * This knowledge is necessary for many applications.


For routing I don’t think it’s very important, an application that comes to mind is in jurisdictions like Germany where you have the obligation to use a cycleway (i.e. you cannot use the road). For rendering it isn’t important to know which way the sidewalk refers to, it is more important to know it is a sidewalk at all.
From a practical point of view, people have often been omitting properties like “name” on sidewalks and cycletracks, which made it hard to give good turn by turn instructions, but this is not inherent to mapping sidewalks separately.

Cheers,
Martin
_______________________________________________
talk mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/talk