[OSM-talk] keys

[Immanuel Scholz]

Hi,

> I think the users (us) are finding that the lack of any structure is
> becoming a bit of a problem. See previous discussions about UK centric
> schemes, highway=autobahn, etc.  I think this initiative is not about some
> complex programming changes but more about the users doing some
> self-organising using the tools they have already got.  I really like the
> lack of any *enforced* structure, but I also see that there are benefits
> to several users sharing a common scheme - that is all that is happening.
>


> > But: Some of your ideas were thought before ;-).
> Where?

The usual channels. Mailing list, wiki, IRC.. I mean - coders already
thought about possible needs of organizing the keys a bit.


>> To have a "registry", where some metainformation about the data is stored
>> was part of the key/value design from beginning. The idea was that
>> keywords can have keywords and so clients are able to read out these
>> keywords and interprete them by themself.
>
> Can you give an example? I don't understand what you mean by "keywords can
> have keywords".

Currently, only Nodes, Segments, Ways (and Areas) are "first class
objects", which mean, you can attach key/values only to this.

It was an original idea to give the user the possibility to attach
key/values to Keys. (Ok, the REALLY original idea was only to have
'keywords' and to be able to attach keywords to keywords. But it comes up
to the same).

Having this, a user can attach for example the key/value
"possible_values=major,minor,highway" to "class". An other example is to
attach "class_icon=http://foo.bar/baz.jpg" to the key "church"

A client can then read out this values and decide to pre-fill a combobox
which selects from a "class" with the values "major", "minor" and
"highway". Futher, the client could display the URL for every node which
has the class=church.


To summarize, the "key/values for keys" is just a simple way of using
existing API's to implement a *distribution* of ontology. It is not better
in actually defining that ontology than any other scheme you can quickly
come up with ;-).




Some things to keep in mind:

- You will never have *no* misinterpretation. I strongly suggest than you
do not make applications that depend on "no wrong data" to work except:
- If you really make an application that has to enforce some scheme (e.g.
routing applications), then you should always add a proxy which filter out
and preprocess your data (as gpsdrive does).


Ciao, Imi