Proposal talk:New place values

From OpenStreetMap Wiki
Jump to navigation Jump to search

Population is the wrong basis

The key as proposed is not useful. For cities in general the population is known and can be looked up and entered into the database (e.g. from Wikipedia). What would be useful is regional importance. E.g. in Germany A city with 50.000 inhabitants that is inside the direct aglomeration of a big city (>400.000)would have to have a rank=-1 or rank=-2 (depending on the model), while a city with 50.000 inhabitants inside the middle of nowhere (like Schwarzwald) with no big cities around, having a hospital, and so on, would have to be classified rank=+2. That's the way it is done in conventional maps. It would be best if this scaling were on a worldwide basis. There needs to be a way for mapmakers to show big cities in lowly populated regions earlier, while showing cities in densely populated regions later.--Extremecarver 08:52, 20 January 2011 (UTC)

I like this version of the rank tag using relative values. We can use the existing place=* values and modify the significance of the place using a rank tag if necessary. A village in a thinly populated area such as the Australian outback can then be tagged as relative important despite a low population since it is the only significant settlement in the area. Depending on the render it can then be shown at lower zooms than otherwise. Negative rank values could e.g. be useful in the famous cases of small towns getting a place=city status for historic reasons where a negative rank could prevent such places from getting an unjustified promenence on the map. I don't find the original proposal using rank=0/10/20/30 workable. It is not intuitive and the mapper will probably have to look up every time which value he should use for this particular population. The values will also be cumbersome to evaluate by the renders. polderrunner 20:52, 20 January 2011 (UTC)
I also like the idea expressed above and originally mentioned by Fredrik Ramm at talk:de, but I don't see it as an alternative to this proposal: this proposal aims to differentiate and extend the existing place-values in a non-harming way (if a data consumer doesn't understand the rank-tag, it doesn't matter or harm, but if it understands it, they can use it to achieve finer granularity), as it doesn't change the existing place-values (like it would be with new values like large_town or similar). The above should be expressed differently like Frederik suggested: rank_modifier=* which is more literal and doesn't collide with this proposal. I agree that it would be useful as well. -- Dieterdreist 10:43, 21 January 2011 (UTC)
The proposal is in my eyes intelligent, but far from reality. I don't think any major map, does not yet account for population to further refine the city categories. Going by very easily controllable facts (like population, or capital vs city) is simply not needed. It is no more work to implement a direct classification based on population, than to enter rank=population into a renderer. Hence as proposed no renderer will use it (as it does not add real information). Regional importance (whereas a region can be the world, Europe or also only a county) on the other hand would need loads of computing power, intelligent algos and many facts that make it really difficulty (e.g. you want to render a map of Germany, but a German extract would not be enough, as for proper analysis you would need maybe 200km bordering region to make it useful, sometimes more (like a big river that only has a short path in one country, but is of continent wide importance).
Still even regional_importance won't be enough. In reality we would need different represantion levels of detail also in map data. There is simply some information that maps show only in low resolution. Like representing mountain ranges like the Alps, or huge forests that in OSM are maybe split into several thousand parts. If you are in low resolution you don't want to see holes inside a forest, even though they exist in reality. Mapmaking means simplification, and every professional mapdatabase but OSM uses simplified models for low resolution. Therefore I don't really see a rank key as really important. It hinders us from attacking the problem at the proper place--Extremecarver 12:36, 21 January 2011 (UTC)

Using population is wrong in most cases

First of all, if "rank is intended to define the general importance of a place", so why places with more population have lesser values of rank? Are negative values possible? --Zverik 19:26, 8 February 2011 (UTC) lesser

I don't see why places with more population have lesser values of rank. Where do you read this? Negative values are not possible. Rank should not be based on population, though population somehow matters (a very small place will never be able to get to the highest rank, the most important places automatically become big). -- Dieterdreist 00:34, 21 February 2011 (UTC)

Also, you assume that place values are assigned based on population. In many countries it isn't so. Usually mappers use either official place status or some complex value based on multitude of factors. Population has its own tag. You suggest to introduce... I still don't get it, but one of two:

  1. a duplicate of population tag, with less clarity
  2. some abstract value, to calculate which a mapper must take into consideration all surrounding places and other factors

If it's the latter, ranking on a scale from 0 to 30 (and more?) does not help: even movies and games do not get more than 10 points. Having big numbers means we must have an algorithm transforming a fixed set of values to the single number which is rank. So why not just put those values into tags and forget ranking? --Zverik 19:25, 8 February 2011 (UTC)

no, I don't assume that place values are directly assigned by population (well, some are actually, like isolated_dwelling), and I am one of the advocats for not using population. A town is a town because it is a town, and a village can have more inhabitants and still remain a village --- up to a certain level. If it reaches a very high population it will most likely be promoted to be a town.
The idea of the scale is to set an absolute maximum (0) and three steps "below": 10, 20, 30. These can later (or right now if there is interest in this) be refined (19, 21, ...) to get a finer grain when needed. Movies in the IMDB for instance (to pick up your example) get ratings like 6.78 while I think we can do with integers. Of course you could also transform the values locally and use a scale from 0 to 0.499 or from 10000000 to 300000000000000. --Dieterdreist 00:34, 21 February 2011 (UTC)

Clearly "North America centric" ;)

To say "Tampa" has the same rank as "Berlin" is ridiculous. "Nobody" in Europe/Asia knows "Tampa" and just because it has a metropolitan area? Nuremberg, for example, has a metropolitan area too (3.5 million) and therefor must have the same rank as Berlin or Munich? Sorry, that doesn't fit. Kellerma 03:38, 21 February 2011 (UTC)

I don't know where you get this from. The suggested values for North America (as the others) are a first draw, not the definite list. I thank you for your comment and agree that Berlin should get a higher (i.e. lower number) rank, e.g. 1. On the other hand for the rank of Berlin it is much more important which rank the other German, Polish, Danish and Dutch cities have. The rank of a North American city is disrupt from European cities (besides ranks 0-2 I'd say). Nuremberg is indeed one of the most important German cities and will have a rank inferiour to Berlin and Munich but still quite low. -- Dieterdreist 16:24, 22 February 2011 (UTC)