Here’s a question about the role of the landuse=residential tag: is it still useful to tag an area with it when more precise detail already exists on the map which clearly shows that the area is residential?
For example, this area has every home outlined and tagged with its street address, so it’s clear at a glance that it’s an area which is predominantly residential.
So should I consider the landuse=residential tag to be a placeholder to be used until such time as more detail (as above) is added; or would it in fact still serve a purpose to add a polygon around all of the houses on the map above and tag it as a residential area? (And if so, why?)
From a user perspective, at low zoom levels it’s more user-friendly to see an entire area highlighted as residential, rather than seeing a map with lots of tiny building specks (or no buildings at all, if they are not displayed at that level).
Sometimes it’s hard to see the forest for the trees, even for humans. It’s even harder for automata. If a software application is looking for landuse=residential specifically, it will blissfully ignore any and all houses.
Also consider the endless varieties of maps that can be rendered based on OSM data. Oftentimes, particularly at lower zoom levels, individual buildings are not rendered at all. The whole neighborhood could disappear from the map unless a landuse=residential area is also tagged.
Finally, there could be attributes that apply to the area as a whole (rather than individual buildings), such as the development/subdivision/neighborhood name. Such attributes ought to be attached to a closed way that also has the landuse=residential tag.
Even if you can infer the area as residential, tools may not, and currently can not. In the end (think 10-20 years) it would make sense to cover the whole world with areas tagged with natural=* or landuse=*.
Land use residential is difficult to derive automatically: and is in many ways more useful than lots of building outlines - however attractive they look on a rendered map.
Residential landuse can be used to infer significant properties of ground-sealing, water run-off, local heat-island effects, and aspects of local ecology. When building outlines are also present these properties can be refined (so that different parameters can be used for densely built-up areas ). This sort of data may be very useful for local campaign or conservation groups who often want good quality accessible data, but don’t have the resources of the developers or other interest groups.
I have, recently, tried to create landcover maps based on OSM data which can be directly compared to similar ones prepared for the European Environment Agency, see http://sk53-osm.blogspot.com/2011/01/simulating-urban-atlas-using-osm.html. The availability of landuse data, and particularly residential landuse made this process pretty easy.