[OSM-talk] Relations not always brilliant

Ian Sergeant isergean at hih.com.au
Tue Apr 8 07:58:26 BST 2008


Lester Caine <lester at lsces.co.uk> wrote:

> until there is some UNIQUE way of tagging high level relationships
> consistently, then there seems little point trying to fix fine detail at
the
> lower level. It brings back up the simple problem of producing a unique
list
> of objects in the data. How DO we currently identify all roads in the UK,
so
> that we don't end up with some of the simply silly links that the likes
of
> Autoroute returns when asking for a location.

This doesn't solve your uniqueness problem, with routes, roads, or possibly
anything else.  Route references within a country certainly aren't always
unique.  Ensuring the reference is in the same country doesn't mean you
still won't get silly results.

A relation provides a unique relation id which distinguishes the M1 in
London, from the M1 in Sydney, from the M1 in Melbourne, from the M1 in
Auckland, etc.  This makes each road reference unique, without trying to
predict the way road references work in different places.

The alternative to using a relation is developing a set of heuristics,
using country, location, reference name, connection nodes, etc.  The
question is whether the complexity of the set that would have to be
developed, and handling the exceptions is better than the complexity of
implementing the required relations.

Ian.





More information about the talk mailing list