[Tagging] Fuzzy areas again: should we have them or not?

Joseph Eisenberg joseph.eisenberg at gmail.com
Thu Dec 24 17:03:41 UTC 2020


Re: Glaciers again.

While it's possible for 2 mappers to disagree on the extent of a glacier
(or a river which quickly changes course) due to using different aerial
imagery, this can be fixed by visiting the location in person and to
confirm the edge of the glacier. This would be a 100% verifiable geometry,
because I could visit the same location and check if the location is
correct within the precision of my GPS device. While this might not be
practical for some glaciers in inaccessible places, it is also possible to
verify the geometry next year when new aerial imagery becomes available.

In contrast, better or newer imagery will never make it easier to determine
where the Sierra Nevada ends and the Central Valley begins. Imagery will
also not help you decide which villages are suburbs of Boston or not. Those
boundaries are inherently non-verifiable.

Re: Bays.

You are right to mention that mapping bays as an area is a problem. This
has been discussed extensively at OpenStreetMap-Carto, and it should be
noted that the decision to render labels based on the mapped area size was
made by one person, without agreement by the other maintainers of the map
style.

Several of us think that it is a mistake to ask mappers to map these huge
geometries which are already well represented by the natural=coastline on
the sides which are verifiable. See
https://github.com/gravitystorm/openstreetmap-carto/issues/3634 and
https://github.com/gravitystorm/openstreetmap-carto/pull/3750.

-- Joseph Eisenberg

On Thu, Dec 24, 2020 at 5:37 AM Anders Torger <anders at torger.se> wrote:

> When I made the starting post I didn't have a better name than "fuzzy
> areas", and I don't really have it still, but what I meant to address were
> those areas that are coined as "non-verifiable geometry" in relation to the
> verifiability principle.
>
> Glaciers are really hard to make an accurate outline of. I've done a few
> of those and the problem I have here in Sweden is that you need late summer
> photos to actually see where the ice is under the snow, and you don't get
> that in the free photo layers, and you also need very accurate topology
> correction of the photos, which indeed has become much better in recent
> years but still leaves more to be desired.
>
> However, regardless of that they are generally verifiable geometry in the
> verifiability principle context. The challenge with glaciers is that of
> accuracy not of verifiability.
>
> Bays with wide openings to the sea on the other hand, where does it end?
> That border is undefined and thus non-verifiable, and according to a strict
> interpretation of the verifiability principle, which I see leading OSM
> community members honor, those should then not be mapped as a polygon, but
> should be mapped as a point placed somewhere clearly inside (verifiably
> inside) the bay.
>
> But let's talk about glaciers again, and other features whose challenge is
> accuracy not verifiability. These features can lead to "edit wars" if you
> have two mappers that use different sources and each of them believe theirs
> is more accurate. Why do I bring that up? Because the risk of "edit wars"
> due to conflict of how the geometry should be drawn seems to be the primary
> argument against allowing drawing of non-verifiable geometry. The argument
> goes that there would be lots of edit war so we would need a policing
> entity and by that we would through the whole egalitarian cooperation
> principle out the window.
>
> However my answer to that is that the danger is exaggerated and we already
> have plenty of non-verifiable geometry all over the map, without any big
> issues of edit wars. They are not really much more danger than those
> features that are hard to draw accurately. And we can not just leave
> everything that out. Oh well, we can if we police it, but we're not really
> that active in that (vandalism is more important to stifle) and we can
> clearly see that mappers want to map these things, so I don't think we
> should outlaw it, but instead drop the passive resistance and let it evolve.
>
> It does add value to the data. With an approximate geometry available a
> renderer can make proper decisions of how big to make a text label, and
> it's also free to move it around a bit to avoid collisions and improve
> readability. Geometry also makes queries better, if you ask the map to show
> a named bay or a plateau or forest or whatever, it can zoom to that place
> and show the proper area of the map where it is rather than zooming in max
> on a point. It shouldn't draw the outline though, as it's defined as
> "fuzzy". Queries like asking if feature X borders to feature Y also
> requires that geometry is available. Sure there can be inaccuracies if you
> ask things like "is feature X within fuzzy feature Y". Looking through the
> history there are suggestions of advanced features where you can define how
> wide the fuzzy border is so the map could answer "unclear" on such a
> question in addition to true/false, but I think that is overkill. I would
> say that inaccurate answer on border queries is a cheap price to pay.
>
> It's more about if we *want* these geometries to exist or not. If we want
> them, we can have them and their issues can be managed. If we don't want
> them we can look at their specific challenges and exaggerate and just say
> no to any proposal. But it seems like we will have some of them anyway
> (they are in), unless we start policing and removing.
>
> The problem with status quo is that there's mixed messages to mappers,
> these geometries are in there and some are rendered, and are obviously
> useful to the alternative (mapping as a point), but they are still
> according to some against the rules, so when someone wants to develop this
> further with a new tag, say for a plateau, it is at great risk of getting
> roadblocked by the same strict interpretation of the verifiability
> principle. Status quo also means a lot of passive resistance towards these,
> for example by not rendering them, like peninsula and valley tags today.
> There's one thing not rendering snowmobile paths and other speciality
> information, another to leave out generic map information like name of
> valleys, to me as a mapper that is a message that we really don't care
> about this type of general-purpose information or at least not if it must
> be made with a "non-verifiable geometry".
>
> That lead me to write the somewhat confrontational heading "should we have
> them or not?" because I as a mapper is really tired of these mixed
> messages, and in my personal mapping I already use these geometries a lot
> and I am in need of more tags (plateau etc) and like others to be rendered
> (valley, peninsula...). If powerful community members intends to backtrack
> and move back to point mapping or even just continue the passive resistance
> it means that a major motivational factor to map is swept away for me, that
> of being able to properly name nature. So it would be nice to know where
> OSM is heading...
>
> A sidetrack of this discussion is if this should be in a different
> database or not. It seems like GIS experts think it should be separate, but
> OSM is not a traditional GIS system. I'm for any solution that works
> though, but it must work seamlessly with OSM editors though as these
> geometries are plentiful and needs crowd sourcing by the mappers just as
> all other geometries.
>
> On 2020-12-23 23:41, Graeme Fitzpatrick wrote:
>
>
>
> On Thu, 24 Dec 2020 at 03:27, Martin Søndergaard <sondergaard246 at gmail.com>
> wrote:
>
> most interesting discussion I have seen on this list yet. And I want to
> give a few of my thoughts as well.
>
>
> Thank you, Martin!
>
> Excellent post & I agree entirely with what you say.
>
> As I mentioned a while back (may have been in another of Ander's threads?)
> - we draw an area & say that this is a town / village, but what about that
> house 300m down the road, & the one 500m beyond that? Do those people think
> they live in this town?
>
> My own city is a major built-up area, but as you go out into the
> surrounding country, you come to suburbs with acre / <hectare house blocks,
> then a bit further there are multi acre / hectare blocks, but where does
> the "city" end?
>
>
> On Thu, 24 Dec 2020 at 03:58, Joseph Eisenberg <joseph.eisenberg at gmail.com>
> wrote:
>
>
>    - natural=glacier with ~56,000 entries
>    - place=archipelago with ~1,300 entries
>
> These three are not at all fuzzy, in the mathematical sense.
>
>
> I'd question a couple of those thanks, Joseph.
>
> Glaciers are constantly moving, so how can they not be fuzzy, especially
> now with climate change apparently melting Arctic glaciers at an ever
> increasing rate, so their boundaries must be constantly changing?
>
> & with archipelagos, & bigger island nations, whose boundaries are drawn
> across the open ocean - how can we be precise about that line?
>
> Thanks
>
> Graeme
>
>
> _______________________________________________
> Tagging mailing list
> Tagging at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/tagging
>
>
> _______________________________________________
> Tagging mailing list
> Tagging at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/tagging
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20201224/ee9484f4/attachment-0001.htm>


More information about the Tagging mailing list