|
Sometimes I see that there's redundant (to my mind) data on the map. For example, there's a building with filled Also, there're several amenities inside this building. And each amenity also has filled Should such things be fixed as long as all amenities have address of the building? |
|
In general address information on POIs is a harmless duplication that in some situations is even useful. Tthere is no real reason to remove it. Though it's not harmless, but is it really useful? Maybe it is useful if the building has several addresses, POI should have their own address tags then just to choose correct from a few.
(03 Nov '17, 09:29)
Sergey Karavay
Assume one of the POIs moves down the street, if you move the node and change the address that is documented. The other issue is, as you note, that the building could have multiple addresses or the address could be on an entrance node, which again would make it sensible to have the address information on the POIs. Note: I'm somewhat a convert in this respect, I used to tends towards reducing such redundancy, but now think it is OK (I won't actively add it in simple cases, but do not remove it).
(03 Nov '17, 09:48)
SimonPoole ♦
|
|
OpenStreetMap does not treat addresses as objects in themselves: they are usually attached to other elements such as buildings, land parcels or points of interest and other amenities. The fundamental nature of the OSM data model does not provide for non-geographically locatable information, and thus modelling addresses as objects is not realistically possible*. Therefore duplication of address information is inevitable and not something to worry about. (Note that most components of an address (street, city etc.) will always be duplicated in OSM. For information on "addresses as objects" papers and slides by Morten Lind on the Danish address system, such as this one, are a good start. On a practical level, data consumers may only be interested in a subset of POIs. Removing address data from such POIs will make it harder to consume such data. At a theoretical level, the OSM data model does not enforce and does not expect normalisation of tagged data. This is true of most free-format tagging approaches or folksonomies. * Relations in principle allow anything; so it is, theoretically. a valid approach. 2
This answer makes good arguments, but it needs to be pointed out that this is not a resolved topic and different opinions on the matter do co-exist. In particular, a common expectation found in mapping practice is that addresses are "inherited": If a building has an address, then all entrances of the building, as well as the shops and amenities inside, share that address by default. Likewise, if a site (e.g. a school grounds) is tagged with an address, then all buildings on that site share the address unless otherwise specified. And while there are practical benefits of duplication, there are also downsides: It can easily lead to the same address showing up on a map multiple times. So neither solution is clearly better than the other. Be careful and use communication when updating other mappers' work by either adding or removing redundant address tags.
(04 Nov '17, 15:18)
Tordanik
@Tordanik, actually, that was my point when I asked. I also stick to inheritance. POIs can always fallback to parent's (building's) address when they don't have their own.
(04 Nov '17, 16:46)
Sergey Karavay
@Tordanik this issue is actually a serious topic in formal database design. It is unlikely that one can treat the issues in a Q&A such as this. It is true that in the vast majority of cases addresses can be inherited, but there are plenty of 'black swans' to show that this is not always the case. But the simple example of a post office box is a useful counterfactual to common assumptions about address data.
(04 Nov '17, 22:00)
SK53 ♦
|
