[OSM-talk-ie] [OSM-talk] All places in Ireland - GNS Data now accessible!

Dermot McNally dermotm at gmail.com
Sat Sep 15 11:21:45 BST 2007

On the matter of redundancy - the is_in field isn't a great place to
store your definitive expression of what county a place is is, as it
is designed to allow inclusion of multiple properties (so Cabra could
be "in" Dublin (city), County Dublin, Leinster and even Ireland,
depending on what other mappers chose to add in). This is why I
figured that a dedicated field that wouldn't be subject to change
(except correction of the wrong ones).

On correcting the county errors - my view is, get them on the map and
clean them up once they get there. Once the points are all placed
creative people may be able to derive county boundary lines from them
(as Christian's own GNU-Plot map did), and this will quickly point us
to anomalies.

On the source date, I'm not inclined to map that to any OSM property
(does one even exist in common use?). Again, this is because any date
stored in this way will be assumed to apply to the full set of OSM
"native" attributes, whereas we can readily assume that these will
evolve based on input from other mappers. Population, county
information and even position (see below) are likely to change. The
GNS date would remain available in the gns: attribute namespace.

On 15/09/2007, Brent Easton <b.easton at exemail.com.au> wrote:
> >Anything's got to be better than the GNS approach of "include the same
> >POI two or possibly three times in different languages apparently at
> >random", though I guess this relates to the quality of the source
> >material used by the NGA. It would be interesting to know the copyright
> >status of this source material.

My approach has been to merge the name information available from the
multiple entries that many places have (the unique ID allows this) and
to derive my best guess at English, Irish and common (usually English)
names. On copyright, in the absence of information to the contrary,
I'm prepared to believe that the distributors of these data sets have
the right to allow us the unrestricted use that they do.

> Be aware that this data is typically recorded to the nearest minute of lat/lon and so you are looking at a typical mean error of roughly 1 km on any given point. You will find that all the POI's will line up into a 1 minute square grid (which will be rectangular except on the equator).

That would be annoying, but probably still useful subject to a cleanup
operation. I can test your theory, as I'm inclined to import these via
JOSM rather than direct API calls. This will allow us to compare data
with existing reality.


More information about the Talk-ie mailing list