[Openstreetmap-dev] OSM - Schema - Phase 1 - Request for Comment

Andy Robinson Andy_J_Robinson at blueyonder.co.uk
Tue Nov 29 14:38:56 GMT 2005


Imi wrote:

>You (as an user) want to simple add data features as "points of interest"
>and "streets" or even "power line", correct?

and all those other beautiful map features sure :)

>
>But under the hood, it is already thought and discussed about how to store
>these different datas in the database.

I wholly appreciate that.

>The wiki page you mentioned in your original posting is not for the users.
>It is not the list of types you are limited to enter in a client. It is
>the list of data types that are used under the hood.

I understand that too. But I think it helps when the two at least have a
degree of commonality for the discussion. It certainly helps me to
distinguish between the point, line and area features on a map. 

I am not suggesting at all there is a need to change the data model as I
think all of the root elements I listed are already covered, albeit that
they are perhaps not yet being utilised (eg "streets" and areas).

>The clients (as example the applet) may provide you (the user) with much
>more features than the database has.

Accepted. The database will always be limited to its physical content. What
the end user does with that content is wholly a separate issue.

>Ok, here we go. My view:
>
>I think that for the programmers, the datatypes mentioned in the wiki are
>enough: keys, nodes, segments, streets and areas. Given the powerfull (and
>simple) property-system, the clients can provide things like "point of
>interest", "bus line", "marketplace" or
>"look-where-I-walked-through-the-park route" without adding more
>complexity to the server.

That's very encouraging to hear.

>In the first place, the access to this additional stuff will be by
>entering specific key/value pairs, as example setting "class=poi;" to a
>node or so.

>This has to be defined soon!
>http://www.openstreetmap.org/wiki/OpenStreetMap_Ontology could be a place
>to start this (which is a broken link from
>http://www.openstreetmap.org/wiki/index.php/Key_Value_System).

That was the actual purpose of starting this discussion rather than to
question the database model. I wanted a structure for the next phase to help
the users respond to the key/value suggestions by placing, for instance,
those relating to liner features under the <way> element.

>Later on, when OSM has more experience with the key/value system and when
>some common key/value pairs has been established, I predict that clients
>will add shortcuts to "create a point of interest" without entering some
>arbitrary "foo=bar;" stuff.

And I believe we should try to narrow down some of the common key/value
pairs as part of the extension of this discussion. Would that help focus the
code writers, like yourself, on what the users need?

>But the need for special data types like POI should not make the core code
>more complex if it is not really necessary.

I was under the impression the POI was already one of the key data
primitives but I see you've just edited that out slyly from the wiki ;)
If the DB <node> table/fields are capable of holding any data relating to a
point, Point of Interest, or node then I agree these are all the same under
the hood. That's confusing for the end user though, where the distinction
between gps points, points of interest and nodes which are used to create
poly features needs to be made, at least for explanation and discussion
purposes.

Thanks for the feedback.

Cheers,

Andy





More information about the dev mailing list