[OSM-dev] curves

[bvh]

On Thu, Jan 03, 2008 at 05:53:08PM +0100, Frederik Ramm wrote:
> >> What people think about explicitly supporting curves as a part of  
> >> geometry model? This way we can model the world better.
> > When I suggested this before, someone said it was patented, possibly by
> > either NavTeq or TeleAtlas, 
> Navteq seem to have a patent on using Bezier curves for road network 
> representation or something. It was on the lists here.

When we would not have beziers in the data, but instead a tag that hints
that the points should be interpolated to beziers, surely that is not
patented?

By not directly introducing beziers but only allowing for 'interpolated
points set', we will never have to deal with 'difficult' beziers
(ie where the line between the endpoints is radically different
from the bezier)

> I also find the idea very tempting, however one has to bear in mind that 
> some things could then become very difficult. For example, computing the 
> combined length of all motorways is trivial today, and will probably 
> require specialist tools if we go Bezier. Equally, finding intersections 

Well, the length along straight lines would provide a good estimate.
For tools that need to be really accurate, the extra code is not _that_
much (10-20 lines of code)

> and the like (think Validator plugin) - probably a computational 
> nightmare if Beziers are involved.

Intersection of beziers can get tricky, but using the straight lines
will be enough for the validator plugin. Again due to the fact that
the interpolation process (when carefully chose) will not generate
beziers that have their control points way of from the line between
the endpoints.

> Will the built-in PostGIS stuff ("give me everything that intersects 
> with this rectangle") work out-of-the box with Beziers?

Will it need to?

cu bart