[OSM-talk] relations in order not to
David Ebling
dave_ebling at yahoo.co.uk
Tue Feb 12 07:42:36 GMT 2008
> As far as I understand it, the idea is simply to
> qualify a tag with
> start and end node. I.e. you have a way that goes
> from node A, B, C to
> Z, but from B to D and from M to P it is a
> pedestrian road. So,
>
> old scheme:
>
> split way into 5 parts (3 non-pedestrian, 2
> pedestrian) and tag
> accordingly.
>
> scheme with "superway" relation:
>
> split way into 5 parts and create one relation to
> contain them all;
> add all common tags to relation; add pedestrian tag
> to 2 ways.
>
> scheme with "qualified tag" relations:
>
> do not split way. create two relations that each
> contain the way, plus
> the start and end node (B/D for relation 1 and M/P
> for relation 2),
> plus the special tag (pedestrian).
Am I the only one reading this discussion thinking
that editing on OSM is going to get so complicated
that we'll have very few new contributors, and
certainly not many non-techies?
I agree that segmenting roads is not ideal, but I'm
just trying to think about the way that relation data
will be presented to users for editing with either of
these two models. Perhaps that's cart before horse,
but it worries me.
I started playing with OSM just after segments had
been done away with, and sometimes I wonder why that
happened, not having ever used them. It seems to me
that the current proposals regarding ways/relations
are somewhat similar to segments/ways. The "qualified
tag" approach seems like ways become what was
segments, except they cover more than one node, and
the relations become what was ways... or am I
misunderstanding somewhat? I feel that the term
"relation" for something that is basically a
meta-data-carrying way will be confusing for
newcomers. The use of "relation" for saying that two
ways are related makes more sense.
Sorry for the rambling stream-of-conciousness, anyway!
Dave
__________________________________________________________
Sent from Yahoo! Mail - a smarter inbox http://uk.mail.yahoo.com
More information about the talk
mailing list