[OSM-talk] Zero tolerance on imports

Frank Steggink steggink at steggink.org
Sun Feb 20 09:14:05 GMT 2011


On 11-02-20 01:03 AM, Richard Fairhurst wrote:
> This is getting crazy.
> Exhibit 1:
> http://twitter.com/#!/maproomblog/status/39053538692698112
> "Whoever imported CanVec in Aylmer, Quebec obliterated hours of work 
> and introduced hundreds of errors. #osm #openstreetmap #whybother"
Of course this should never have happened. I absolutely agree with 
Jonathan and you about that.
> Once again, some keyboard jockey has decided that his l337 import 
> skills are better than the knowledge and hours of work by a local 
> mapper. The offender appears to be user 'sammuell' by the look of it - 
> http://www.openstreetmap.org/user/sammuell - though he hasn't posted 
> anything about his activities on the user page, the wiki, or indeed 
> anywhere.
I don't know where you have looked, but apparently you didn't do a very 
good job. The user you're pointing out is importing Canvec and making a 
lot of contributions in Manitoba, which is a long way from Aylmer / 
Ottawa. Some excuses would be appropriate.
> This is killing OSM. We are not here to provide a free API to 
> government geodata that can be obtained trivially elsewhere. OSM is 
> all about "added value"; by deleting genuine surveyed data in favour 
> of mindless duplication of other, poorer quality datasets, we are 
> _destroying_ value.
How about the reverse case? I've seen that many times. Sometimes it's 
really obvious that data is not "genuinely surveyed", but just pulled 
out of thin air.

I'll confess, I'm an importer. I have imported and am still importing 
data in the Netherlands and Québec. Not in Aylmer though, but mostly 
Central Québec. I've spent countless hours in surveying the city of 
Québec itself and the (major) highways in Central Québec myself. Because 
the scale of the province and the entire country, it is simply 
impossible to depend on local users for contributing to OSM. During my 
time I only had contact with a handful of other locals, of which I only 
one met (she happened to be a colleague). The other users I met came 
mostly from the Sherbrooke area, which is over 2 hours away by car.

Anyways, because there are only a few active users, I decided to join 
the Geobase NRN (roads) efforts at one point, and Canvec at the 
beginning of this month. With Geobase I did my proper duty, by joining 
up the roads to the existing ones. With Canvec I omitted all roads. 
There are names on them which are not on Geobase, but I'll look at that 
later. Regarding the age of the data (which can be 30 years old): it is 
still better than nothing for an area which is too vast to survey by 
individuals like you and me. Also keep in mind that there is only 
Landsat coverage, except for the larger urban areas. So, I think that it 
is appropriate to use as a template on which other users can elaborate.

The import in the Netherlands is landuse only (the infamous 3dShapes 
import, for the statisticians among you). We're trying to engage other 
users too, although this isn't always going smoothly. In the end we 
always came to some kind of agreement, so that's a big win for OSM. (No, 
3dShapes doesn't always draw the long straw.) After an import of a sheet 
we always do a post-import cleanup. That means evaluating 
inconsistencies, conflicts, etc. Since the import is landuse only, I 
restrict the cleanup mostly to this, although I do some minor cleanup of 
obvious errors, like self-intersecting ways, untagged nodes, etc., as 
well. Many of the pre-existing data which gets cleaned up comes from the 
AND import. In other cases I'm using Bing imagery as the judge. 
Sometimes I keep the imported feature, sometimes the user feature, 
sometimes both.

In case anyone is completely up in arms about this: this is why I can 
tell that user contributed data is not always better than imported data. 
(And imported data is also better than no data, unless the imported data 
is plainly wrong and misleading). It is not as black-and-white as you 
paint. So, please lower your pitchforks, and accept that imported data 
CAN BE better than user-contributed data. OSM has over 360k registered 
accounts, and it is impossible to keep everyone happy. Nearly everyone 
is genuinely doing his best, and I'm pretty sure the user who messed up 
in Aylmer did this by accident. He should have been contacted first, 
before the entire anti-import brigade is stumbling over him. Anyways, 
now I will have that coming, by writing this mail :)

Importers are humans too, and all humans do make mistake. However, I do 
realize that the work (and especially the mistakes) of importers are 
much more scrutinized. This is OK, because of the impact of their work 
in certain areas is much larger.
> From what I can tell (talk-ca postings etc.) 'sammuell' is a fairly 
> inexperienced OSMer who presumably thinks "this is how things are 
> done". It isn't. How do we stop this impression taking hold? How do we 
> explain that imports are _not_ welcome except as a last resort, and if 
> you do them, you _must_ follow a very, very rigorous set of guidelines?
Richard, you're just one of the 360k contributors, and one of the 
millions of users. I do acknowledge that you've made a major difference 
into making OSM what it is today. However, that doesn't mean that your 
opinion can be hold as the absolute truth. I believe the average 
community opinion is more like: imports _are_ welcome, but _only_ if 
there are no better alternatives, and _only_ if a strict set of 
guidelines is followed (for example _not_ deleting better quality user 
contributed data). It's most beneficial for OSM when the best of both 
worlds are combined.
> cheers
> Richard
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk



More information about the talk mailing list