[talk-au] Suburb boundaries - getting close

Jack Burton jack at saosce.com.au
Mon Feb 16 11:41:52 GMT 2009

On Mon, 2009-02-16 at 22:09 +1100, Franc Carter wrote:
> Ok, it seems my conversion script is now producing sane results so
> it's time
> to work out what the final output should look like.
> The first question that I think we need to answer is, how do we
> represent the
> data in OSM, there appears to be 3 options:-
>    1. Closed ways
>    2. Relations
>    3. Borders with a left/right tag

My vote would be for option 1. I won't bore the list by repeating all
the reasons for that.

> Then we need to decide on what tags to apply to the data. The raw data
> has three fields
>   * STATE_2006     A numerical identifier for the state the suburb is
> in
>   * SSC_2006        An identifier provided by the ABS
>   * NAME_2006      The name of the suburb, which may have the old name
> in '()' after it.
> So, my initial proposal for tags is:-
>   * name=?
> (with any old name removed)
>   * source=Based_on_Australian_Bureau_of_Statistics _data     (ABS ask
> for this)            
>   * ABS:reviewed=no
>   * ABS:STATE_2006=?
>   * ABS:NAME_2006=?
>   * ABS:SSC_2006=?

Suggest that where there's an old name, put it in an old_name= tag.
Then, with the current name in name= and the old name in old_name=,
there'd be no need to keep the ABS:NAME_2006 field.

Similarly, it would make sense to put the contents of the STATE_2006
field in an is_in= tag, preferably with ",Australia" appended. If we're
going to keep this field in the dataset, it may as well be in a format
that's consistent with what's already used.

ABS:reviewed=no & ABS:SSC_2006=foo would both seem essential, for
tracking changes, as would the source= tag to satisfy the license.

> The 'ABS' part is just a suggestion - It's a bit short for my liking

Yeah, "ABS" does sound a bit short, but then again,
"Australian_Bureau_of_Statistics" sounds kinda long. How about
"abs.gov.au"? - short enough to type in one breath (e.g. for search) but
still uniquely identifiable.

> We also need to decide where these tags go - nodes, ways, relations.

I think tagging the nodes is a bit pointless, with only three possible
exceptions: (a) if the dataset includes central nodes for each suburb,
in which case I'd suggest tagging those, but not any other node; (b) if
the attribution requirement of the license mandates attribution on
everything (in which case tagging nodes with just a source= tag should
suffice); or (c) if you plan to update the dataset in some sort of
automated manner in the future (in which case the nodes probably only
need the ABS:SSC_2006=foo tag) - but from earlier discussions on the
list, this does not seem to be the case.

If you go with option 1 or 3, then obviously the ways need to be tagged.

Under option 2, I'd suggest tagging both the relations and the ways, so
the data can be used in as many different manners as possible.

Just my 2c worth...



More information about the Talk-au mailing list