Hi,

John Smith wrote:
> You are confusing things,

[...]

> So a straight line in the database as on the planet will still be a
> straight line,

We were discussing what exactly a straight line was. There is no such 
thing as a "straight line in the database", because, as you correctly 
state, the database only stores the end points of a line. If you draw a 
line from point lat=10;lon=10 to lat=30;lon=30, then it is unclear 
whether that line visits point lat=20;lon=20. Some might think yes, some 
might think no.

* If you have drawn this line in an editor from a satellite imagery 
background and your editor displayed things in Mercator, then the line 
that you saw on the screen when you entered the two end points will 
*not* go through that point.

* If your editor was using EPSG:4326 then the line you saw on the screen 
*will* go through that point.

That is the problem we are talking about, not...

> if the rendering software needs more points it needs to
> deal with it.

... simple two-dimensional line drawing algorithms. Unless you add extra 
tags to the way which tell the renderer how to interpolate between the 
two points, there is not enough information to know.

Now either we provide that information, by making a rule and hoping 
everyone understands and adheres to it (unlikely), or else we just try 
and keep our nodes close enough to each other because that will then 
reduce the error introduced by the ambiguity discussed above, to 
something that we do not have to care about.

Also, just in case that has not become clear enough already, our "map" 
API call does not catch lines that intersect the bounding box without 
nodes in between, so any editor/renderer relying on this API call will 
not even get a chance to "deal with it" because the software won't even 
know that there is a line to be drawn. This is another reason why 
keeping your node distances in the < 5km range makes a lot of sense.

I hope this has made things clearer for you.

Bye
Frederik

_______________________________________________
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk

Reply via email to