<br><br><div class="gmail_quote">On Wed, Oct 20, 2010 at 11:20 AM, Peter Budny <span dir="ltr"><<a href="mailto:peterb@gatech.edu">peterb@gatech.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">"Andrew S. J. Sawyer" <<a href="mailto:assawyer@gmail.com">assawyer@gmail.com</a>> writes:<br>
<br>
> My thoughts are mixed in below.<br>
><br>
> On Wed, Oct 20, 2010 at 11:17, Peter Budny <<a href="mailto:peterb@gatech.edu">peterb@gatech.edu</a>> wrote:<br>
><br>
> Antony Pegg <<a href="mailto:antthelimey@gmail.com">antthelimey@gmail.com</a>> writes:<br>
><br>
> > tagging admin area / populated centers / labels in USA seems to<br>
> > come down to two main tags:<br>
> ><br>
> > admin_level and place<br>
><br>
> Before you over-simplify, let me point out a couple things:<br>
><br>
> 1. Not all of the US is incorporated. In the Northeast, every tiny part<br>
> of land is incorporated into a town or township or borough. But in the<br>
> Southeast (and I presume elsewhere as well), there's lots of<br>
> unincorporated land, even in the vicinity of large cities. Look at<br>
> Atlanta, which still has lots of unincorporated area.<br>
><br>
> That's a big variation, and the map needs to be equally competent at<br>
> handling both regions.<br>
><br>
> Slight correction, not all land in the Northeast is incorporated. In New<br>
> Hampshire there are a handful of communities which are not<br>
> incorporated.<br>
<br>
</div>I was exaggerating to illustrate the differences, but point taken.<br>
<div class="im">><br>
> 2. Defining how "important" a city is (and thus, how big its label on<br>
> the map should be) is a tricky thing to do. Population is certainly a<br>
> large factor, but how do you define this? The City of Atlanta is the<br>
> #33 most populous city in the US, with 540,000 people, but the Atlanta<br>
> metropolitan area is #9 with 5,475,000 people and is the largest metro<br>
> area in 800 miles.<br>
><br>
> There's also a recognition factor... the whole world knows where New<br>
> York is and would expect it to be fairly prominent on a map. Capitol<br>
> cities are considered to be "important" even when they're not very<br>
> prominent or populous. Etc.<br>
><br>
> It seems to me that admin_level handles the first point, except that 4<br>
> levels to cover all of the US doesn't give much granularity. Maybe we<br>
> need to think about using the in-between levels to show more detail?<br>
><br>
> place= seems to be handling the second point, but not very well. Should<br>
> label sizes really be determined purely by population? By "importance"?<br>
> What criteria should there be? I don't think the current scheme of<br>
> city/town/whatever is very good, because it's another instance of<br>
> hacking a British scheme onto a country with a very different history<br>
> and organization.<br>
><br>
</div><div class="im">> I agree that there isn't a one-size-fits-all approach that will work<br>
> with displaying/tagging named communities on the map. I think that a<br>
> combination of the size of the given area, the "admin_level" of the<br>
> given area (country, state, county, etc), population and<br>
> recognizability (capital cities, etc). The latter being the most<br>
> difficult to quantify in a manner in which many people would agree on<br>
> (less capital cities).<br>
<br>
</div>I forgot to mention control cities (<a href="http://en.wikipedia.org/wiki/Control_city" target="_blank">http://en.wikipedia.org/wiki/Control_city</a>).<br>
These are cities that are designated for use on highway signs to<br>
indicate which direction you're heading. These should definitely appear<br>
on the map, even if they're relatively small cities (e.g. Valdosta,<br>
Georgia).<br>
<div class="im"><br>
> However, I agree that a ratio of area, "admin_level" and population<br>
> could take care of most cases.<br>
<br>
</div>This gets me wondering if maybe there's some way to do it more<br>
automatically. For instance, it should be easy to find data sources<br>
for population, area, and lists of "global cities"<br>
(<a href="http://en.wikipedia.org/wiki/Global_city" target="_blank">http://en.wikipedia.org/wiki/Global_city</a>) and control cities. Maybe<br>
there should be a process just before the renderer that takes in that<br>
information and decides how to label cities automatically. That keeps<br>
the OSM database down to the basics.<br>
<font color="#888888">--<br>
</font><div class="im">Peter Budny \<br>
Georgia Tech \<br>
CS PhD student \<br>
<br>
</div><div><div></div><div class="h5"><br></div></div></blockquote><div><br></div><div>If you haven't seen it, the 41Latitude blog had an earlier post about getting a good density of US city names at different zoom levels--it was in reference to Bing maps, but the same ideas apply: <a href="http://www.41latitude.com/post/931787074/improving-bing-maps-3">http://www.41latitude.com/post/931787074/improving-bing-maps-3</a></div>
<div><br></div><div>In his analysis, using the Census' Urban Clusters and Micropolitan Statistical Areas looked useful. Worth a read, if you're still following this thread. :)</div></div><br>