Hi Pierre,
I'm glad to see you're taking a constructive approach towards the
discussion initiated by Paul :) I definitely agree something needs to be
done to make imports more manageable, especially the Canvec import.
There is also an amount of "work left behind" as I mentioned in my other
mail. Of course that needs to have some attention as well.
Apart from the inconsistencies, there is for example also the Geobase
import. In 2009 the available Geobase data didn't contain road names in
Québec, and by then it was also not clear when it would contain names.
So there are many areas which have roads without names. Furthermore,
there have been subsequent Canvec releases in which data / tagging
issues have been resolved. An example is the use of "natural=land" tag
as mentioned by Brian Gamberg, but in one or two releases the
amenity=school and amenity=prison tags were swapped around as well.
Having some kind of checklist attached to the micro-tasking, combined
with periodical reviews (which IMHO think are necessary, for example
with new Canvec releases) and tools like OSM Inspector will also ensure
that the data quality remains good in the future. For this system to
work, also areas with existing data (either user-contributed or
imported), which seem "complete" should be reviewed.
I'm not familiar with the Linz solution, but perhaps someone like
Martijn van Exel or the Mapbox guys could help setting up something
useful and user-friendly.
Frank
On 13-11-2012 21:01, Pierre Béland wrote:
Voir discussion en français / See English discussion below
Paul Norman,sat. november 2012 6h37
** Subject : Internal CanVec conflicts
> Is it the communities view that it is okay to import CanVec without
> reconciling the internal differences between the layers?
> My view is that importing data without resolving conflicts of this type
> where it conflicts with either existing data or internally is not an
> acceptable import and indicates the importer did not sufficiently review
> what they were uploading.
---------------------------------------------------------------------------------------------------
L'import de données est essentiel, si l'on veut couvrir toute la
surface du Canada. Cependant, il est complexe d'importer des fichiers
CanVec dans les zones où les données existent déjà. Autant les
contributeurs inexpérimentés qu'expérimentés sont susceptibles de
faire des erreurs. Le processus d'importation est souvent trop
complexe et trop long à réaliser.
Les micro-tâches sont actuellement basées sur les zones géographiques
,chaquegrille NTS étant subdivisée en plus petites zones. En
subdivisant par couche thématique, je pense que cela permettrait de
réduire la complexité des importations CanVec, de réduire les erreurs
et d'encourager plus de gens à importer. Si nécessaire en raison de
la taille, certaines grilles pourraient aussi être subdivisée en zones
plus petites.
Tout comme pour les fichiers Planet, les fichiers d'import OSM
pourraient être subsiviés par couches telles que routes, poi, landuse,
forest, coastlines, limites administratives, autres ...). De cette
façon, chaque tâche d'importation serait moins complexe à réaliser,
plus facile à comparer avec ce qui existe déjà. En outre, la tâche
serait réalisée plus rapidement.
Lorsque une couche telle que les forêts semble moins approprié pour
une région donnée, il serait facile pour le contributeur d'ignorer
cette couche. Aussi, les limites administratives et les côtes
devraient être réservés à des gens plus expérimentés. Les fichiers
pourraient être regroupés dans un répertoire distinct et couvrir de
plus grandes zones.
Je pense que nous avons besoin de plus que le fichier Google doc
actuel pour assurer le suivi des imports.
L'outil linz2osm de Nouvelle-Zélande me semble trop complexe.
Cependant, il peut nous donner des idées sur la façon de développer
un tel outil.
Voir linz2osm New Zealand project.
http://linz2osm.openstreetmap.org.nz/
voir discussion de Glen Barnes, Import list.
http://web.archiveorange.com/archive/v/2u2n5O1bELI3yg2ULjry
---------------------------------------------------------------------------------------------------
Data import is essential to cover all of Canada, But it is complex to
import Canvec files in areas were data already exist. Both
unexperienced and experienced people may make errors. Import process
is often too complex and too long to realize.
Micro-tasking presently consist of dividing a a NTS grid area in
smaller zones. If this micro-tasking was based on layers, I think that
this would reduce the complexity of Canvec imports, reduce errors and
encourage more people to import. But if necessary because of size,
some NTS grids could be subdivided by smaller zones.
The OSM import files would be divided by layers like it is done for
planet files. There could be layers such as roads, poi, landuse,
water, forest, coastlines, administrative boundaries, other...). This
way, the individual import tasks would be less complex to realize,
easier to compare with what already exist. Also, the task would be
realized more rapidly.
When certain type of data such as forests seems less appropriate for a
specific area, it would be easy for the mapper to skip this layer.
Also, the administrative boundaries and coastlines should be reserved
to more experienced people. They could be grouped in a distinct
directory and cover larger zones.
I think that we need more than the Google doc to monitor the mapping.
The New Zealand linz2osm tool seems too complex to me. But it can give
us some clue about how to develop such a tool.
See linz2osm New Zealand project.
http://linz2osm.openstreetmap.org.nz/
See Glen Barnes discussion on Import list.
http://web.archiveorange.com/archive/v/2u2n5O1bELI3yg2ULjry
Pierre
_______________________________________________
Talk-ca mailing list
Talk-ca@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-ca
_______________________________________________
Talk-ca mailing list
Talk-ca@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-ca
