I’ve recently been swayed to the persuasion that landuse features should be considered more as a separate entity than other elements in OSM.
The non-standard nature of the current state of landuse mapping means that some mappers map every single plot as a residential area, others map a city block as a residential area, and some mappers map an entire suburb as a single residential area. This spectrum of different ways these objects are mapped in my opinion makes them impractical to try and use for anything other than as a tool to paint the map and signify the purpose the land is used for.
Because of this it’s easier to make a dataset of all mapped McDonald’s when they’re all mapped as a single node rather than when some of them are tagged onto buildings, and some are tagged on a landuse area. I think that it isn’t a good idea to guide mappers to map POI data onto landuse, or to put names or other metadata on any landuse if there’s a viable alternative.
With residential areas, there are already viable alternatives to tagging the name of a neighborhood or community than on the residential area. That’s where the place=* key comes in. you can use tags such as place=neighborhood or place=hamlet, and they are much more appropriate and easier to query objects than landuse=residential.
The problem arises with commercial and retail areas. If you’re tagging a shopping center or office park, your only option is to put the name of the shopping center or office park on the landuse area, since there are no alternative tags to put the name on.
For this reason I’m considering making a proposal to introduce new tagging for this problem, something along the lines of place=offices or place=retail to represent these objects with an easier to process data scheme, but I’d like some public input first.
Hi SherbetS,
I understand your concern about different and non-homogeneous ways of mapping landuse, but I definitely don’t agree on the general conclusion that it equates to only “painting” the map. Landuse information is very relevant for many purposes, even if not homogeneous, and many other features are mapped differently in different areas of the world.
Having said that, on the more specific consideration about naming some landuse areas (and specifically commercial and retail areas) I don’t see the real problem in assigning a name to them: could you specify what bothers you?
It seems to me that a proposal for using a place-based tag wouldn’t be good, because you’re not really naming the place after the commercial company or activity… but maybe I’m missing your point here.
I’m saying that we should move away from the model of generally encouraging users to use landuse to represent POI data, because it can be difficult for a data consumer to process landuse data correctly, and increases the risk of a mapper mapping it in a weird or unexpected way.
I’m suggesting using the place tag to signify the name of an office park, or the name of a retail center, not as the name of any particular office or store, so that we can rely less on landuse as the primary way to map a named area.
Thanks, I’ve understood better your point of view.
Anyway, if an area is used as a commercial or retail area, it is not only acceptable, but desirable that that area is identified with a landuse=commercial/retail tag. If a mapper has the possibility to add information about the activity, she/he can add it to the landuse, a specific building inside the area or a node. I don’t really see the need to use the place key for that.
Furthermore, it seems to me that this would be a misuse of that key, which should be used to identify a location known by a particular name (Key:place - OpenStreetMap Wiki), namely “every significant human settlements (city, town, suburb, etc.) and also for notable unpopulated, named places”, and not a POI identifying commercial or retail activities.
I think that it’s less preferable to put the data on the landuse area, because of the reason I mentioned before that different mappers map landuse differently, making it very difficult to determine exactly what a landuse area represents. Having a place POI node to represent the office or retail area would be appropriate (IMO this counts within notable unpopulated named places)
If there’s a better key to represent this object, I’d be happy to consider it.
-SherbetS
the term “area” can have several meaning, I would prefer if we say “residential landuse” rather than area, because “residential area” is also used for neighbourhoods or city quarters (i.e. place entities) with supposedly predominant residential use. This is already sorted in OSM, people should use place=neighbourhood / quarter / suburb for this kind of residential area. Landuse is about the usage of land, not settlement parts with names. There is no difference when you split a landuse=residential polygon into 2 landuse=residential polygons with the same extent, “landuse” is a property, it is not a countable feature.
Generally, my suggestion is to map landuse on small polygons and by excluding roads, because it is easier for everybody, the one drawing it, the next mapper who has to do some changes in the area (e.g. add a different landuse) and because the additive nature makes it unlikely that things become unnecessarily complex, and it also leads to precise mapping as areas with different usage are excluded from the beginning, and you can see where people have actually looked at the landuse and where the map is still “blank” in terms of landuse.
The opposite approach, mapping the whole village or suburb or at least huge chunks, as residential use, leads to many different areas not being mapped (wrongly attributed residential landuse) and when these areas are finally mapped, it often leads to more and more complex multipolygon relations because of all the inners, something which is hardly every necessary in mapping built up landuse (this is all refering to landuse in settlements and not or less to landuse in the open, like forests and fields). On the plus side, this mapping style leads sooner to a map that is seemingly complete, and without further processing, it may be generally more appealing visually, particularly in medium and low zoom levels. But this comes at the cost of rough and imprecise landuse maps which tend to be harder to maintain and be less inviting to adding more details / refinements.
The problem arises with commercial and retail areas. If you’re tagging a shopping center or office park, your only option is to put the name of the shopping center or office park on the landuse area, since there are no alternative tags to put the name on.
why can’t you put place=neighbourhood on an office park? We already have a tag for shopping centers (shop=mall) so there may be less pressure but neighbourhood could apply there as well
Anyway, if an area is used as a commercial or retail area, it is not only acceptable, but desirable that that area is identified with a landuse=commercial/retail tag. If a mapper has the possibility to add information about the activity, she/he can add it to the landuse, a specific building inside the area or a node. I don’t really see the need to use the place key for that.
if you tie the name to a landuse object it means you cannot map different landuse within without excluding those things from the named object.
On the place=neighborhood wiki page it says:
Don’t use neighbourhood for uninhabited places or isolated settlements
I think this is an issue because we’re encouraging users to simultaneously use the landuse=commercial tag on just general commercial areas that don’t have names, as well as for named areas.
Having a place POI node to represent the office or retail area would be appropriate (IMO this counts within notable unpopulated named places)
a polygon is preferable to a node for such kind of typically very clearly delimited areas as office parks. It adds a lot of implicit information like extent and size, shape and orientation.
Don’t use neighbourhood for uninhabited places or isolated settlements
we can remove the part that requires residences, it never was part of the definition that was approved by voting. I’d agree for isolated settlements, as these are usually hamlets or villages.
In English, neighbourhood directly implies somewhere where people live in a community, it doesn’t make sense to use it on an unpopulated place like an office park.
a polygon is preferable to a node for such kind of typically very clearly delimited areas as office parks. It adds a lot of implicit information like extent and size, shape and orientation.
I’m on the fence about using smaller place=* areas on an area, but if they have an authoritative boundary I think it’d be okay.
that’s what was already written on the page though: “The tag can be used for any kind of landuse or mix of landuse (such as residential, commercial, industrial etc). ”
it is also in line with similar place values, which are really just code words not to be taken literally in any aspect, a place=suburb may well be in a central part of a city while the meaning of suburb in natural language is sub-urbs, it is literally “below city”, in the low density outskirts.
I think it is sufficient that mappers agree on the limits, authoritative boundaries are mapped with the boundary tag. Place is for socio geographic places which often are not officially defined
Yes, we do not need landuse for shops and offices. A node or a line with the usual tags is enough (e.g. shop with more than one building)
But we don’t need anything new either.
Have you already discussed this topic somewhere in OSM? I remember that someone was claiming some time ago that tagging name on landuse=retail is bad but I do not remember where this disussion has taken place.
Can you link examples where current tagging is not enough to solve that?
Not really. For some purposes merging POIs with buildings is annoying, but for just listing location mix of all mentioned methods is fine.
The problem is that mappers are using landuse=retail objects to store name data. this is better represented with a point for similar reasons that we use place=neighborhood, etc. for populated settlements.
If the area has a name, such as with a shopping centre, then add it using name=*.
I’m not sure I follow. It is not better to use a point. Even for place=neighbourhood, we prefer polygons to mere nodes. The main reason why we still sometimes use inferior nodes for e.g. neighborhoods, is when the exact boundaries are unclear / fuzzy, so we cannot really make a precise polygon - e.g. building in one neighborhood will look exactly like the nearby building in another neighborhood - the only difference is administrative, and not something visible/verifiable on the ground. See neighbourhood#Node_or_Area wiki for details.
However, for the landuse such confusion should not be a problem - i.e. you can usually clearly tell on the ground if some buildings are supermarkets or apartment buildings, so there is no need to fall back to mere nodes, as superior polygons can be used instead.
It seems to me that you find polygons as worse than nodes, which I would disagree with.
Can you explain why you think that (if that is indeed what you think), or try to clarify what I missed/misunderstood?