[talk-ch] Larger removal of street and place names derived from the GWR

marc marc marc_marc_irc at hotmail.com
Sun Aug 13 14:55:29 CEST 2017


Le 05. 08. 17 à 13:00, Simon Poole a écrit :

>> 117 OSM Roads 107 GWR Roads
>> 100 OSM Roads 98 GWR Roads
>> I doubt that the municipality has removed 9 roads. What happened ?
> No, that is the difference between what had names -before- the redaction
> and -after- so 9 names have been removed from OSM (not the roads
> themselves).

107 -> 98 GWR Roads : I thought it means that 9 names have been removed 
in the GWR database (independently of the revert in osm) and ask what 
can be the reason for this decrease of GWR Roads in GWR database for a 
municipality that hasn't announced any destroyed/merged routes ?

> the number of errors is likely to be small).
I make a query to quantify my feeling about the quality of the revert

In Fribourg, way last modified by dwg since 2017-08-01
1163 have now no name -> easy to survey
221 still have a (supposed wrong) name -> what do you think if I add a 
fixme "name altered by DWG. survey needed" ? it 'll help to warn where a 
new survey is needed to fix a typo or a wrong road limit.

In Domdidier half of the roads are involved at least partly.
3/5 of the affected road with a wrong name have been survey by myself 
AFTER the problematic import. so I suppose that survey changeset done 
after the import are not taken into account.
1/5 of the affected road have now a correct name so the deleted name 
wasn't imported from GWR.
so IMHO in Domdidier, the revert hit ratio is 20% problematic data 
removed <> 60% survey data destroyed <> 20% not imported data removed
IMHO I find that an error rate 4 times higher than what was problematic 
is not as small as you think.
I'm of course unable to do this at the whole country scale.

A much more important issue : what todo in the future to prevent a 
verified information by survey from being deleted because someone made a 
mistake a few months/years ago ?
Because if the named of affected road was not set or wrong, I would have 
added/fix myself when I survey the whole road.
Should we backup all changeset before upload ?
It would be problematic to replay them a few months or years later 
without conflicting with the changes made in the meantime by the others,

> I had a quick look at what stolenaddresses.io is doing
> wrt Fribourg and they are scraping a non-public arcgis
> server to generate the dataset.
Thanks for the check. I 'll make another threat in a few day about 
opendata done by Bern and how try go the same for Fribourg.


More information about the talk-ch mailing list