[OSM-dev] [OSM-talk] Osmosis and large bounding polygons

[Brett Henderson]

Martijn van Oosterhout wrote:
> On Jan 9, 2008 10:58 AM, Frederik Ramm <frederik at remote.org> wrote:
>   
>> Sounds good, especially as I have the simplification code lying
>> around, in Java, somewhere; only thing that bugs me a bit is that
>> this would do the simplification every time you use the polygon which
>> strikes me as unneccessary. Haven't tried the 5 MB polygon people are
>> talking about, but I guess it will take some time for this to be
>> simplified.
>>     
>
> My question is: why is the polygon so complex? Yes, the
> northern/southern borders of the US might be a bit complex for some
> parts, but the east/west coast would be a few dozen segments at most.
> Australia I think you can do in maybe 50.
>
> The problem with automatically simplifying is that you don't know
> what's important. For example, when cutting out NL for the AND import
> it needed to be *exact*. But to have it done in a reasonable time
> someone simplfied the border in areas where there was *no data*. Also
> the entire coastline was replaced by 10 segments. That cut it down
> enourmously but is not something that can be automated.
>
> So maybe it could be an option but it would be better if people worked
> out whether the polygons really needed to be complex in the first
> place...
>   
That's a good point.  I guess when people are playing with something for 
their own purposes it's easy to hit maproom and grab the polygons, a 
simplify method would be useful in this case.  Any serious usage (such 
as AND import) would require some further investigation.  Australia in 
particular as an island can be represented as a very rough polygon, 
presumably you just have to be careful to include or exclude islands as 
appropriate.

In any case, the default for the --bounding-polygon task would be 
simplify=no so it wouldn't be enabled unless somebody wanted it.

One thing I've been curious about is coastline data.  Presumably a 
polygon is likely to include coastline in some places and exclude in 
others depending on accuracy.

I'll add this to the list but it's unlikely to reach the top for a while.