[Talk-ca] RIP CanVec
Il y a actuellement un peu de bruit sur talk.ca pour faire renaître les importations CanVec. Il ne faut pas. Les importations CanVec (et les importateurs CanVec) sont responsables de plusieurs problèmes: - ils ont fait fuir plusieurs contributeurs OSM (explications plus loin). - des données désuètes ont remplacé des données à jour. - des données imprécises ont remplacé des données précises. - CanVec a introduit des erreurs de modélisation en imposant dans OSM le concept de tuile plutôt qu'en respectant celui de l'objet. - CanVec a un mauvais traitement des objets chemin (ou rue). Il n'y a pas de nom (du moins au Québec) et il y a un segment pour chaque pâté de maison. Cela n'a pas de sens dans plus de 90% des cas. Les importateurs CanVec n'ont pas suivi les guidelines OSM et ont agi en solitaire. Ils ont agi en utilisant la stratégie tabula rasa et ont effacé tout un territoire pour faire place à leur importation. Des dizaines (centaines, milliers?) de contributeurs en ont été frustrés et ont cessé de collaborer à OSM. Les importateurs qui n'ont pas fait tabula rasa ont créé un problème de duplicata. Des lacs existants ont maintenant un doublon dont la qualité est douteuse. On n'a qu'à regarder dans les Laurentides. Un lac situé sur une fontière de tuiles se verra découpé en 2, 3 ou 4 parties. Ce lac sera défini pas les zones inner de plusieurs relations et il devient difficile de fusionner ces entités volumineuses. Deux cas possibles: - on ne fait rien et on laisse la carte dans un piteux état. - on détruit les méga-relations de CanVec et on recommence. Dans les 2 cas, la valeur ajoutée par CanVec est discutable. Mon intention n'est pas d'affirmer que CanVec n'a que des torts. Grâce à CanVec, nous avons: - plus de rues - les adresses (addr:interpolation) - la forêt (natural:wood) Malheureusement: - la façon dont CanVec traite les rues demande beaucoup de travail aux contributeurs qui veulent fusionner tous les segments de rues. - les lacs ne devraient pas faire partie des relations de forêt car les renderer d'OSM ont pour règle que les lacs sont au-dessus de la couche de base. Mes recommandations aux importateurs CanVec (si CanVec renaît): - n'importer que dans les zones où il n'y a pas de contributeur (Nunavut par exemple) - impliquer les contributeurs s'il y en a - laisser les rues et routes aux contributeurs OSM - bravo pour les adresses - ne pas tenir compte des lacs ou des chemins dans les contours de tuiles - ne pas importer un lac s'il existe déjà dans OSM. Il est probable qu'il a été dessiné à partir d'un imagerie (Bing, Yahoo, ...) qui est plus précise que que les données CanVec faites à partir des données d'un satellite désuet. Pour le dynamisme et la pérennité d'OSM dega ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] RIP CanVec
On 2014-Nov-17, at 1:53 PM, Ga Delap gade...@gmail.com wrote: Il y a actuellement un peu de bruit sur talk.ca pour faire renaître les importations CanVec. Il ne faut pas. Unfortunately, though I can read French, I can only speak or write English properly, but wanted to weigh in on this discussion. As someone who has imported a fair bit of Canvec data, I wanted to weigh in on this. Though Canvec data has some issues involved in importing into OSM that people should be aware of, I think that it does far more good than harm if treated properly. All data from Canvec (or any source for that matter) should be inspected carefully before importing into OSM - using satellite imagery for verification, for instance. And in particular, any data from Canvec that is replacing existing data in OSM should be considered very carefully and with a high degree of suspicion, as in most cases (though not all), our existing OSM data is superior. But with that said, there are a lot of pieces of data from Canvec that we don't have in OSM and should be added - more streets, addresses, and forest areas, as you mentioned - especially if they're verified by overlaying satellite imagery. I'd add to that list hydrography. In particular, smaller streams and lakes in my part of Canada tend not to exist at all in OSM, and adding them from Canvec adds data where there otherwise was none. It's usually best to leave the larger lakes and rivers that already exist alone, though perhaps adding a bit more detail to their shorelines. In my experience, probably about 90% of the data in Canvec (particularly non man-made features) that is not already in OSM may be worth importing. And similarly, about 90% of the data that is common between the two is best left alone. Some areas of the map in more remote areas don't have any data at all in OSM - in those cases, importing Canvec data particularly adds a huge amount of value. In places like downtown Toronto, probably not so much. I think Canvec data is very valuable for OSM and very strongly support its continued importation, under the condition that whomever is doing it knows some of the issues as you pointed out to ensure they are always making things better instead of worse. Dan --- Dan Charrois President, Syzygy Research Technology Phone: 780-961-2213 ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] RIP CanVec
I don't think the two of you are actually disagreeing at all. Ga Delap just said that the past importers of CanVec data were too haphazard and that there was not enough community consulting. It seems like the quality of the data in Quebec might have been worse than in other places too. (Désolé, je ne parle pas francais bien.) Je ne pense pas que vous differez. Ga Delap a dit seulment que les ancients importateurs de donne CanVec etiez trop négligé est ils n'agissent pas avec la communauté. Il me semble aussi que la qualite de donne quebequois est pire que les autres. On Mon, Nov 17, 2014 at 2:00 PM, Dan Charrois d...@syz.com wrote: On 2014-Nov-17, at 1:53 PM, Ga Delap gade...@gmail.com wrote: Il y a actuellement un peu de bruit sur talk.ca pour faire renaître les importations CanVec. Il ne faut pas. Unfortunately, though I can read French, I can only speak or write English properly, but wanted to weigh in on this discussion. As someone who has imported a fair bit of Canvec data, I wanted to weigh in on this. Though Canvec data has some issues involved in importing into OSM that people should be aware of, I think that it does far more good than harm if treated properly. All data from Canvec (or any source for that matter) should be inspected carefully before importing into OSM - using satellite imagery for verification, for instance. And in particular, any data from Canvec that is replacing existing data in OSM should be considered very carefully and with a high degree of suspicion, as in most cases (though not all), our existing OSM data is superior. But with that said, there are a lot of pieces of data from Canvec that we don't have in OSM and should be added - more streets, addresses, and forest areas, as you mentioned - especially if they're verified by overlaying satellite imagery. I'd add to that list hydrography. In particular, smaller streams and lakes in my part of Canada tend not to exist at all in OSM, and adding them from Canvec adds data where there otherwise was none. It's usually best to leave the larger lakes and rivers that already exist alone, though perhaps adding a bit more detail to their shorelines. In my experience, probably about 90% of the data in Canvec (particularly non man-made features) that is not already in OSM may be worth importing. And similarly, about 90% of the data that is common between the two is best left alone. Some areas of the map in more remote areas don't have any data at all in OSM - in those cases, importing Canvec data particularly adds a huge amount of value. In places like downtown Toronto, probably not so much. I think Canvec data is very valuable for OSM and very strongly support its continued importation, under the condition that whomever is doing it knows some of the issues as you pointed out to ensure they are always making things better instead of worse. Dan --- Dan Charrois President, Syzygy Research Technology Phone: 780-961-2213 ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
[Talk-ca] Fwd: RIP CanVec
Oops, meant to reply to the list, not an individual. :) -- Forwarded message -- From: Adam Martin s.adam.mar...@gmail.com Date: Mon, Nov 17, 2014 at 7:43 PM Subject: Re: [Talk-ca] RIP CanVec To: Alan Trick alantr...@gmail.com I have had some experience regarding the use of Canvec data. I have not tried to import this data myself nor would I try - it is complex enough that I think it is better that I keep my fingers out of it. My experience with Canvec has been limited to correcting the data in areas that I am editing in. I echo the statements of Ga Delep in part - the data, in some areas, has been haphazardly imported with features out of alignment and possible destruction of user contributed data. But on the balance, I think it has been a good information source in the hands of experienced importers. Land features, such as local hills, groves, open landscapes, reefs ... these are excellent pieces of information for the map. It has it's issues - the tiled nature of the data is one problem. Another is the enormous multipolygon relations that result from imports of these areas - they are difficult to modify and correct because of the large size. They might be inaccurate when one compares them to the fine details, but they serve the purpose when there is no other data available. Honestly, I destroy these polygons when I encounter them AND I am in the process of correcting the data where they exist Ensuring the data remains available to these handful of skilled individuals is a good idea, as is ensuring that the community is consulted with the data is being imported. On Mon, Nov 17, 2014 at 7:27 PM, Alan Trick alantr...@gmail.com wrote: I don't think the two of you are actually disagreeing at all. Ga Delap just said that the past importers of CanVec data were too haphazard and that there was not enough community consulting. It seems like the quality of the data in Quebec might have been worse than in other places too. (Désolé, je ne parle pas francais bien.) Je ne pense pas que vous differez. Ga Delap a dit seulment que les ancients importateurs de donne CanVec etiez trop négligé est ils n'agissent pas avec la communauté. Il me semble aussi que la qualite de donne quebequois est pire que les autres. On Mon, Nov 17, 2014 at 2:00 PM, Dan Charrois d...@syz.com wrote: On 2014-Nov-17, at 1:53 PM, Ga Delap gade...@gmail.com wrote: Il y a actuellement un peu de bruit sur talk.ca pour faire renaître les importations CanVec. Il ne faut pas. Unfortunately, though I can read French, I can only speak or write English properly, but wanted to weigh in on this discussion. As someone who has imported a fair bit of Canvec data, I wanted to weigh in on this. Though Canvec data has some issues involved in importing into OSM that people should be aware of, I think that it does far more good than harm if treated properly. All data from Canvec (or any source for that matter) should be inspected carefully before importing into OSM - using satellite imagery for verification, for instance. And in particular, any data from Canvec that is replacing existing data in OSM should be considered very carefully and with a high degree of suspicion, as in most cases (though not all), our existing OSM data is superior. But with that said, there are a lot of pieces of data from Canvec that we don't have in OSM and should be added - more streets, addresses, and forest areas, as you mentioned - especially if they're verified by overlaying satellite imagery. I'd add to that list hydrography. In particular, smaller streams and lakes in my part of Canada tend not to exist at all in OSM, and adding them from Canvec adds data where there otherwise was none. It's usually best to leave the larger lakes and rivers that already exist alone, though perhaps adding a bit more detail to their shorelines. In my experience, probably about 90% of the data in Canvec (particularly non man-made features) that is not already in OSM may be worth importing. And similarly, about 90% of the data that is common between the two is best left alone. Some areas of the map in more remote areas don't have any data at all in OSM - in those cases, importing Canvec data particularly adds a huge amount of value. In places like downtown Toronto, probably not so much. I think Canvec data is very valuable for OSM and very strongly support its continued importation, under the condition that whomever is doing it knows some of the issues as you pointed out to ensure they are always making things better instead of worse. Dan --- Dan Charrois President, Syzygy Research Technology Phone: 780-961-2213 ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org
Re: [Talk-ca] RIP CanVec
On 14-11-17 03:53 PM, Ga Delap wrote: - CanVec a introduit des erreurs de modélisation en imposant dans OSM le concept de tuile plutôt qu'en respectant celui de l'objet. Amen to that. Is there any way to de-tile the data? I realise that most of Canada is one giant water relation, but is there a data processing pipeline that can recognize and join up split entities? I'd also be up for running a simplification routine on every way imported (or reimported). While ways can have a maximum number of nodes, they all don't need to have that number, and the original CanVec points aren't sacrosanct. Unless these issues can be carefully addressed, I'm firmly in the “Il ne faut pas” camp. cheers, Stewart ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] RIP CanVec
I'm near Lloydminster, Alberta and there is lots of data missing in the area that we use OSM for.I use Canvec to fill in missing data in my area. Canvec data should not be used to override more precise data but there are Geobase imports in my area that are a complete wreck. Some of the roads are so bad that it's significantly less time consuming to just erase the road and reimport versus trying to fix it. I do spend some time with my gps and look for corrections while out driving but in our area we'd have significantly less use for OSM without the Canvec data. Darren Wiebe On Mon, Nov 17, 2014 at 6:50 PM, Stewart C. Russell scr...@gmail.com wrote: On 14-11-17 03:53 PM, Ga Delap wrote: - CanVec a introduit des erreurs de modélisation en imposant dans OSM le concept de tuile plutôt qu'en respectant celui de l'objet. Amen to that. Is there any way to de-tile the data? I realise that most of Canada is one giant water relation, but is there a data processing pipeline that can recognize and join up split entities? I'd also be up for running a simplification routine on every way imported (or reimported). While ways can have a maximum number of nodes, they all don't need to have that number, and the original CanVec points aren't sacrosanct. Unless these issues can be carefully addressed, I'm firmly in the “Il ne faut pas” camp. cheers, Stewart ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] RIP CanVec
On 11/17/2014 5:50 PM, Stewart C. Russell wrote: Is there any way to de-tile the data? I realise that most of Canada is one giant water relation, but is there a data processing pipeline that can recognize and join up split entities? I looked at this, but it's better to go back to the original sources, particularly now that CanVec is no longer being updated. ___ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca