[OSM-talk] Abandon all ways: a polemic

Laurence Penney lorp at lorp.org
Tue Mar 20 17:20:03 GMT 2007

Raphaël Jacquot wrote:
> the definition of a way is typically in the GIS world
> * goes from one intersection to another intersection *
> whoever says "the common definitions from the GIS world don't apply to
> us because <blah>" is mistaken.
> they represent topology at the most lower level, that is you can use
> that way to go from a point to another point without intesecting with
> another way.
> now, this definition implies that this may be different from a street.

That would be lovely! If ways represented topological edges (one
intersection to the next) then they would certainly represent
something useful, notably for route-finding.

Please forgive me if this is what the whole "ways as node sequences"
is all about. It's a good plan, equivalent to the segments-only
approach except there's an arbitrariness in that a way of 10 nodes can
be represented as any number of ways between 1 and 9; and also some
ways will need splitting if they go beyond a topological edge.

I'd simply prefer to have a model where those two problems above
simply cannot exist. The post-processing in the segments-only approach
would be to compress the attribute data, and to cache the topological
edges; neither of these would change data as far as a human tagger
could see.

> a street is composed of a number of ways, and ways composing that street
> may have different navigation rules applied to them.
> for instance, a street may be 2-way for most of it's lenght, except for
> a little bit in the middle where it's either one way, forbidden except
> to non motorized vehicle, access requires permission which is handled by
> some device (automatic or manual bollards) or whatever else us, stupid
> humans can think of to piss off other humans.
> now. I'd like a proper explanation of why you feel a way shouldn't have
> a arbitrary number of nodes in it's composition.

When I say arbitrary, I am talking about ways that are longer than
topological edges, such as ways in the db now that represent major
roads: the same way passes minor roads, yet stops & restarts for no
apparent reason (perhaps editing convenience).

-- Laurence

More information about the talk mailing list