[OSM-talk] colour fills in osmarender

Tom Carden tom at tom-carden.co.uk
Fri Jun 9 12:07:07 BST 2006


On 09/06/06, Etienne <80n80n at gmail.com> wrote:
> On 6/9/06, Raphaël Jacquot <sxpert at esitcom.org> wrote:
> > > Are loops and branches in ways a good thing or a bad thing?   I can
> > > imagine that route planning applications might find them problematic,
> >
> > I believe we shouldn't have loops of any sort.
> > routing algorithms like to have straight simple topologies. loops add a
> > huge level of complexity that I don't think is necessary

Absolutely agree.  I think ways as a data type will be pointless if
they allow loops and branches, and I've always said so.

> To do this, ways would need to be constrained so that segments can
> only be contiguous within a way
> (segment/@toNode=following-sibling::segment/@fromNode).  This would be
> a good thing IMHO.

I would like to see any existing loopy or branchy ways in the database
divided where necessary, and the server to enforce non-looping and
non-branching in the API.

> However, if you don't have ways with loops and branches then you don't
> really need segments at all.
>

In terms of UI, I think this is the way to go right now.  If people
want to make a way containing only one segment, that shouldn't be a
problem.  The interface should be geared to
create/merge/manipulate/annotate ways, not segments.  I'd be
interested to hear from Richard what his focus will be for his
proposed Flash editor - will it work this way?

In terms of data model, I still think having an option to just spit
out segments from the API is useful - people will always disagree as
to what the precise definition of a way should be, but segments are
totally intuitive.

> Some questions though:
>
> How would a roundabout be defined?

I would have thought that depends on whether the roundabout has a name
or not, and precisely how the streets cross it.  Mostly, I would have
thought that the loop itself would be a way, unless there is a clear
cut case where one road obviously runs continuously over a roundabout,
in which case the way-ness of that road should take priority and the
roundabout should fill in the gaps with as many ways as are needed.
This will vary from roundabout to roundabout, someone might care to
catalogue the potential configurations and find examples of each, but
that someone probably isn't me (on email... I'd be happy to sketch it
on a beer mat, such are my inclinations).

> Is that a way or a point?

Any serious roundabout should not be a point.  Mini roundabouts that
people routinely drive across should probably be a point.

>
> Some roads, especially residential housing estates, can have roads
> that split and merge and branch, but only have one name.  How can that
> be achived economically?
>

Economically is easy - just represent it with as many ways as is
necessary.  That will be more economical than annotating each segment
individually, which is one of the main justifications for using ways.
Optimally is a different question, but doesn't really matter.

T.




More information about the talk mailing list