[Talk-ca] Aylmer/Hull QC: CanVec import overwriting existing edits

[Frank Steggink]

On 11-02-19 09:27 PM, Jonathan Crowe wrote:
> This is rather discouraging and frustrating.
>
> When the Bing imagery was made available, I spent a fair amount of
> time cleaning up the maps in Aylmer, Quebec, a western suburb of
> Gatineau. Just discovered that my edits have been overwritten by
> another user importing CanVec data.
>
> This import has introduced hundreds of errors:
>
> - Proposed and under-construction streets are shown as completed.
> - Streets are tagged as unclassified.
> - Service roads have been removed altogether.
> - Turning circles and traffic signals are missing.
> - Certain streets do not line up with high-resolution Bing imagery,
> and not in an off-by-a-few-metres way -- for example, the Bing image
> has a subtle S-curve, the CanVec data has a rather crude straight line
> bisecting the curve.
>
> This means that Aylmer now needs dozens of person-hours to get it back
> to where it was -- and at this point I'm discouraged enough to say the
> hell with this project and find something better to do with my spare
> time.
>
> Much of Ottawa is in a similar state: manual edits superimposed with
> CanVec imports that may or may not have been more accurate, and now
> there are two or three duplicates on top of one another. It's a mess,
> and it's in no way ready for MapQuest.
>
> I thought CanVec imports were frowned upon when there was existing road data?
>
> I'll tell you this: this is not the way to encourage people to
> contribute, if all the work we do with imagery is subject to
> obliteration by someone else's work with government vector data.
>
> Yours in frustration,
>
> Jonathan Crowe
> The Map Room: A Weblog About Maps
> http://www.maproomblog.com/
>
> _______________________________________________
> Talk-ca mailing list
> Talk-ca at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-ca
Hi Jonathan,

I'm sorry to hear that this has happened. Imported data should generally 
not replace user-contributed data, unless the quality difference is obvious.

Probably the best way to proceed is that someone who has good experience 
with reverse tools to contact you, as well as the offending importer. It 
is best to identify the changesets of the imports, and revert them. This 
should also restore your contributions, as well as those of other users 
in the area. Perhaps the import is done in multiple phases (as if the 
import is done from a clean slate), so more changesets might need to be 
identified.

In case there is data missing, you and other active users in the area 
could decide whether you guys will take it upon yourselves to add it to 
the map, or import non-duplicate data.

Unfortunately this mistake has been picked up by osm-talk, ensuing in a 
renewed discussion about how "bad" imports are. Imports _themselves_ are 
not bad by themselves. The problem lies with the _individuals doing 
imports_ who are not following proper conduct (like a careful evaluation 
of existing and new data, too lazy to pay attention / clean up, etc.).

I'm saying this as someone who has done both user contributions and 
imports. When importing data, and encountering conflicts (duplicate 
data) during the cleanup stage, I'm always asking myself if I would 
prefer to keep it if I were the user who added it. If unsure, I'll 
consult the Bing imagery as a judge. Of course I'm doing this cleanup 
according to my own guidelines, but I strive to take an objective stance 
as much as possible. This mostly applies to my import work in the 
Netherlands (landuse data), where this situation arises much more often 
than in Central Québec (where I've recently picked up Canvec, for now 
sans roads...).

This process is difficult and very tedious, but I'm glad to say that 
since I started doing large scale imports 1 1/2 years ago (first Geobase 
NRN), I've received way less complaints than positive comments. And in 
nearly all cases we came to some agreement. The only time where this 
isn't the case, happens to be in my backyard, but I'll pick that up with 
Bing some day... At least I managed to convince that user that the 
quality of the Dutch landuse data isn't nearly as bad as he painted. He 
also happened to use a cellphone only to record GPS tracks.

Regards,

Frank