[OSM-talk] Extending the 'geo:' uri scheme: Adding parameter 'osmid'

[Greg Troxel]

stevea <steveaOSM at softworkers.com> writes:

> I'll state even more strongly than Frederik just did: "linking to an
> OSM object by ID and expecting the ID to remain constant is asking for
> trouble" is putting it mildly.  It IS trouble.  All it takes is one
> single change to one single datum and boom, the assumption that doing
> so can work is proven false.  I'll offer to be the first to change an
> ID to do this just on the general principle that it proves this is a
> bad idea.
>
> So, this (linking to an OSM object by ID and expecting the ID to remain constant) is a non-starter.  Right here, right now.

And, the real harm is the seconds-after-links-exist push that people
should not edit in a way that changes IDs.

The IETF should not support referring to unstable IDs because that's bad
engineering.

People who want to refer to osm objects need to

  understand what it means when it's a named way and how the reference
  should behave as the way is split and micromapped

  how to bind by type/name so it's stable in some sensible semantic
  sense, sort of a combo of
    name
    type
    rough coordinates (to disambiguate the above)