[OSM-dev] coastline polygons

Frederik Ramm frederik at remote.org
Tue Feb 13 11:03:14 GMT 2007


Hi,

      don't get me wrong, I lke what you are doing, and the map  
quality increases significantly with water in the right spots.  
However, while fixing the problem on the slippy map level is a good  
thing, it does not absolve us from the task of finding a "proper"  
solution for the long term.

> We can put the coastlines on the
> slippy map without them being in the OSM database, so why duplicate?

For example, I may want to map a shoreline walk. Much easier if I  
have the shoreline in JOSM! And: The slippy map isn't everything. As  
long as we do not have the data in our database, then everyone who  
wants to create beautiful maps will have to access our database AND  
some other source to get shoreline data from.

> Disadvantages:
> * The OSM database is struggling as it is. Coastlines are thousands of
> extra nodes that we don't need.

What is the number of nodes required to map every street, track, and  
path in the whole word? (Which is our ultimate goal, isn't it?)  
Compared to that, I assume that the coastline is peanuts. It may look  
a lot at the moment because our database only covers a tiny fraction  
of existing roads, but in the long run it is negligible.

> * Coastlines are much too large to be represented by a single way - it
> will always be "incomplete" when you download it, and things will  
> break.

So are rivers, motorways, and even boundaries of many lakes and land  
use areas. We will definitely have to find a way to work with this.  
If I download a small bounding box from the server, I *do* want to  
know whether this lies within a giant national park or nature  
reserve, even if the boundaries of that are far out.

> * Having the coastline in multiple ways breaks the topology of the
> system, and requires horrible hacks in renderers to make use of.

As I said, it is the system that is inadequate here.

> * The public coastline data derived from STRM and other places is
> adequate - we can't usefully improve it with our GPSs and wiki-like
> editing methods.

Au contraire. I haven't seen the quality of your current data but the  
automatically imported coastlines that are currently in the OSM  
database do have bugs that can easily by corrected in-system if you  
have the data. It is our philosophy to map what we see and how things  
are, not take somebody else's word (or data) for granted. It is great  
to have their data as a starting point, but the OSM contributor is  
the ultimate authority.

Plus, coastline changes all the time, see Japan or Dubai. While many  
of these changes will probybly go unnoticed at the moment, it is very  
likely that at some future point in time if they create a new "World  
Island" somewhere, an OSM contributor will follow the project closely  
and enter the new islands as they are created. *If* given the chance.

>> A simple rule like "sea is on left" or some
>> such, would be more than sufficient to solve your "main problem"
>
> No it wouldn't. That would only solve the problem if you are rendering
> something that actually had a stretch of coastline in it. If you  
> find a
> completely empty area, you have no idea if it is in the sea or in the
> desert.

You are right but this problem definitely has to be fixed; as I said  
there are many other reasons like large estates, forests, nature  
reserves and so on where you will need a hint whether you're in them  
or not even if your bounding box lies 100% within.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00.09' E008°23.33'






More information about the dev mailing list