[OSM-talk-ie] Everywhere in Ireland - on the map
dermotm at gmail.com
Tue Sep 18 19:45:49 BST 2007
Nobody said "No! don't do it!", so I went ahead and imported all the
place names referenced in my earlier posts. The results are pleasing
where you can see them (JOSM, mostly, since the tiles at home layer isn't
getting many re-renders this weather).
I've done most of the tidyup that's really necessary to avoid
confusion (and particular to ensure that, if duplicate nodes get
deleted, it isn't the new version that disappears), but some remains.
Since the Mapnik layer will be regenerated during tomorrow, it's worth
mentioning what you'll see and how I reckon we should work with it:
Since a goodish number of places had already been placed on the map,
some places are now present twice. (Though I've tidied up, at a guess,
about 70% of such places). The fix for this is clearly to delete one
of them, and this gives rise to a dilemma: pre-existing data has
probably been spotted fairly accurately, either from a surveyed GPS
waypoint or from aerial/satellite photos. As against that, it probably
has only an English-language place name entered and is unlikely to
have any of the additional meta-data available from GNS (county,
sometimes population, frequently Irish name).
The GNS data, OTOH, comes armed with the extra meta-info, but is often
quite poorly-placed. My observation suggests that this isn't as simple
as an over-coarse grid, since some places are far more badly out than
others. The place attribute also stands a good chance of being set to
"hamlet" (see below). The other common problem (IMHO) is where the
GNS-claimed definitive name doesn't accord with the real one on the
ground - it favours Dunleary (as do I, but it's not very widely used),
but it calls Newbridge Droichead Nua and Enfield Innfield.
My suggestion, and what I've been doing, is that we always favour the
imported node over the existing one, as having more complete
meta-information. It is also recognisable as GNS data by its unique
IDs, which can be used to avoid duplication if ever we do another GNS
upload. The changes you'll typically need to make are:
* change place attribute from hamlet to town or village.
* change the name attribute if your local knowledge is more compelling
than the data source. If, in doing so, you are changing the language,
ensure that the name:en or name:ga attribute is populated with the
name you have replaced (unless it's completely wrong).
* Carry over any other useful attributes that may be on the old node.
* Reposition the node if required. I did so in cases where I believed
that the old node was based on better info than the imported one (or
in cases where the imported one was clearly off).
Poor, but good enough IMHO to keep what we've imported. We should make
a point of adopting imported points in our areas and correcting their
positions. Out-of-copyright maps (50 years old or more for OSI) can be
useful to us in this.
Most imported locations, other than the major population centres,
didn't come with enough information to allow their size to be
inferred, and have therefore been tagged as place=hamlet to avoid
making a mess of lower zoom levels. For those places that correspond
to townlands or crossroads, this classification seems OK, and
importing this way ensures that manual touchup of this attribute will
be at a minimum. It still needs to be done, though.
The required cleanup is simple - if you discover a place that you know
or that you can demonstrate to be a town or village, retag it. (Cities
are fixed already, except Kilkenny, which apparently isn't
administered as one)
More information about the Talk-ie