[OSM-talk] Mechanical Edit?

Simon Poole simon at poole.ch
Tue Jul 27 07:30:17 UTC 2021

Am 27.07.2021 um 03:07 schrieb Casper Kersten:
> Dear Frederik Ramm and all other readers,
> > If one person sees a horse and tags it "kow" and then another person 
> comes along and says "we all agree that if something is tagged kow and 
> not cow it is almost certainly wrong" and changes the tagging, without 
> noticing that the thing is a horse and not a cow, then nothing has 
> been gained.
> I strongly disagree with your argument here. Many of the objects I 
> changed were still on version one or two and had been for several 
> years until I came along. Taginfo usage and chaos theory suggest that 
> as long as we don’t take action, the numbers of color=*, 
> building:roof:shape=* and other duplicate tags only grow. With my 
> changesets I am effectively undoing the chaos that has been formed, 
> which I think is a valuable contribution to our ever-growing database.

To be clear this has nothing at all to do with chaos theory, but xkcd 386.

color is actually a good example of the use of a key slowly decreasing 
over time after two peaks in 2013 and 2015 (the later clearly due to an 
import/mechanical edit and 2013 maybe too).

> Adding to this, now the roof shapes, colours and other attributes are 
> more likely to be rendered, and if people who look at rendering 
> software (such as OSMbuildings) see that the attributes of objects in 
> their area are wrong, they are more likely to correct them. Thus, to 
> use your example, a rendered “cow” will attract more mapper and user 
> attention than an unrendered “kow” and is therefore more likely to be 
> updated to the correct “horse”.
> Using ctrl+F to unify tags is something that we should all embrace 
> with the huge database that we are maintaining. Your proposal to wait 
> for local mappers to come along and fix all of the millions of 
> outdated tags there are is not only horribly inefficient, but is also 
> not a solution at all, because most objects are only looked at once or 
> twice by local mappers, who are in many cases also unaware of or 
> uninterested in OSM-wide tagging conventions.
> > If you think that a simple search-and-replace is a good idea – then 
> leave your hands off it and let someone fix it who has the time and 
> energy to do it properly.
> DWG member or not, what counts as “proper fixing” is not for you, but 
> for the community to determine. If it were up to you, then ctrl+F 
> would be removed from JOSM and the community would lose a powerful 
> mapping tool. That mindset is harmful to the development and long-term 
> sustainability of OSM and I strongly disapprove of it.
> As I stated previously, we cannot keep looking at the mess in our 
> project, ignore it and then blissfully continue to micromap the local 
> playground. We need to start showing some agency for the larger issues.
You are massively overstating the effect of typos in tagging, not only 
are the numbers small (and due to the increase in preset based tagging 
decreasing in occurrence),  but further typos tend to simply lead to the 
tag being ignored aka simply having no effect.

Just to give a rough example there are ~470 million buildings in OSM, 
even if all 6'000 color values that you fixed on ways had been on 
buildings (a rough glance would indicate that irl there were far less), 
we are talking about a tiny effect, even if we look just at buildings 
with building:colour tags (less than a million) we are still talking 
less than 1%.


> Best regards
> Casper
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20210727/8ff3fc19/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 495 bytes
Desc: OpenPGP digital signature
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20210727/8ff3fc19/attachment.sig>

More information about the talk mailing list