[OSM-talk] New API suggestion: Allowing contributors to easily track their OSM-objects over time

pangoSE pangose at riseup.net
Sat Aug 22 23:04:51 UTC 2020


Hi Martin :)

Den Sat, 22 Aug 2020 19:55:16 +0200
skrev Re: [OSM-talk] New API suggestion: Allowing contributors to
easily track their OSM-objects over time:

> sent from a phone
> 
> > On 22. Aug 2020, at 19:44, pangoSE <pangose at riseup.net> wrote:
> > 
> > Maybe we should first add permanent ids (new table) and reference
> > that.  
> 
> 
> we do have permanent ids for nodes, ways and relations. ;-)

Okay I think I understand what you are saying. You are saying that if
I create a node, edit it in another changeset it still has the same
id, right? (same goes for the other 2) The problem is that the rest of
the world talks about physical objects like houses, hospitals and
toilets and not nodes, ways or relations and things relying on OSM break
when mappers enrich the map and e.g. embed nodes in a way and move the
tags to the way. 

> What kind of permanent ids do you want? For some more abstract
> concept like a road with a specific name? A shop? A building? If
> there’s a way tagged with building=supermarket, shop=supermarket,
> name=Foo, and you add a permanent id to it. Then the supermarket gets
> bought by someone else and changes to name=Pango? Or the supermarket
> closes. Or it gets torn down and rebuilt by the same operator. Or
> someone detaches the shop tags and adds them to a node inside. What
> happens with the id?

Our datamodel and tools currently does not support keeping a higher
order view of an object and letting the user modify it using our
internal representations at will. Instead we treat objects as like the
don't belong together or relate at all unless they are part of a
relation.

From what I understand an uploaded way converted in JOSM to a relation
get a new osm_relation_id, (asuming I first upload a
node with changeset A, change it to a way by adding another node to it
in changeset B and finally convert it to a relation in JOSM in changeset
C). But I might be wrong about this.

> 
> You can have this with relations, but it (associated street, street)
> did not find a lot of support from the contributors and most mappers
> have decided for themselves that it does not work for them.

Yeah, that was perhaps not the best way to implement a model (the name
implies that it only relates to streets btw.) that handles human
features well over time.

Cheers
pangoSE



More information about the talk mailing list