[OSM-talk] maplint vs. uncontinuous streets?

Dave osm.list at randomjunk.co.uk
Thu Apr 26 10:56:34 BST 2007


On 26/04/07, Frederik Ramm <frederik at remote.org> wrote:
> Andy,
>
> > I'll continue to use the flexibility offered by the current data
> > model, until either the API is changed to prevent it or future
> > developments give me a better mechanism. And along those lines, can
> > people please refrain from saying that such techniques I and others
> > use are "wrong", "incorrect" and so on, and more truthfully suggest
> > that multiple ways are "suggested", "a convention" or even
> > "preferred", if that's what you want to say.
>
> [snip]
> As for the Y-shaped ways, you must remember that internally they are
> stored as a *sequence* of segments, and it is simply not possible to
> express an Y-shaped way as a sequence. So your statement that putting
> every part of an Y-shaped way into the same way "most closely matches
> the situation on the ground" is not really, ehm, correct ;-).
>

I would probably disagree with that assessment!
A way is a sequence of segments, yes, but there is no requirement for
those segments to be connected, or in fact ordered in any logical way.

If you take a way to imply a logical grouping of segments, then it is
a more "correct" representation to combine all segments belonging to
that group in the same Way. Obviously the sequence looses some of it's
meaning, but as it actually never had any, that isn't really a problem
:-)

If you take a way to imply a sequence of connecting segments, then it
is more "correct" to have multiple ways. Ofcourse now you expect the
renderer to figure out that a feature has multiple ways, same for
editors... but tbh this probably has less impact on pretty pictures.

Personally, I know which is easier for me to edit, and for now
computers seem to be still getting faster, so I'll leave the difficult
bits up to the CPU and all those very irritated programmers. Software
engineering would be so much easier if it weren't for all those lazy
users :-)




More information about the talk mailing list