[josm-dev] Geometric calculations and projections

[Matthias Julius]

"Tobias Wendorff" <tobias.wendorff at uni-dortmund.de> writes:

> Am Do, 17.12.2009, 23:40 schrieb Anthony:
>> What's the definition of "rectangle" in non-euclidean geometry anyway?
>
> I can't answer this right now ;-)
>
> But since we've got projections which *are* actually good to show
> rectangles with correct shapes and angles, it could be possible to find a
> suiteable solution.

My point was that Wyoming *is* a rectangle in a Mercator projection.
This is because its borders are simply defined by parallels and
meridians.  Therefore, it would make sense to orthogonalize it using
Mercator.  I suspect that most areas of that scale that one could be
trying to orthogonalize are defined in a similar way.

Objects that are usually the target of actions like orthogonalize or
align in cicle are much smaller (a few hundred meters at most).  There
the use of Mercator would only introduce small errors until one gets
close to the poles.  This is why these commands currently work
reasonably well.  Because most people work with Mercator.  

But, if one selects a different projection most of these geometry
commands produce different results because they use the current
projection.

And this is why I started this thread.  I think it would be better if
Mercator was always used regardless which projection is currently
active.  This would be a simple change in the code.  

Matthias