[OSM-talk] Automatic problem fixing is an automated edit

Andy Townsend ajt1047 at gmail.com
Fri Jul 30 15:35:38 UTC 2021


On 30/07/2021 14:51, Casper Kersten wrote:
> Dear Frederik Ramm,
>
> So the community has to ask for your permission each time we press 
> ctrl+F? That seems a little unrealistic, don't you think?
> ...


Hello Casper,

Perhaps it would help to give a bit of background on some of the 
problems that the DWG sometimes sees as a result of this "automated 
fixing".  An example of one of them is:

http://osm.mapki.com/history/way.php?id=128265146

According to the discussion on 
https://www.openstreetmap.org/changeset/16601166 , that's a steak house 
in Banfora in Burkina Faso.  Unfortunately, it was "corrected" to be a 
fast food place because of the name and then further "improved" to have 
the cuisine and brand tags for the well known fast food place with a 
similar name.  All of the mappers involved thought that they were making 
OSM better by either correcting information or adding more - they were 
all unaware that they were removing the results of the original local 
survey from OSM.

There's absolutely a place for "automated problem detection" in OSM, 
provided that you review the results afterwards and are happy that the 
changes that you're proposing to make really are correct and aren't just 
guesswork.  For example yesterday I uploaded 
https://www.openstreetmap.org/changeset/108841357 via StreetComplete.  
It's an area that I'd updated in OSM before, and I thought that I'd 
added all the path surface info, but StreetComplete found some that I'd 
missed.

More widely, as you've found from the comments at 
https://lists.openstreetmap.org/pipermail/talk/2021-July/thread.html#86817 
and https://www.openstreetmap.org/changeset/108430155 etc., things that 
look "obvious" to you might not seem quite so obvious to other people, 
and may well merit further discussion.  As an aside, your change from 
"color" to "colour" as described (albeit after the event) on 
https://www.openstreetmap.org/changeset/108430155 is much less 
problematic than some others that OSM occasionally sees (assuming that 
all coniferous trees are evergreen was one) - it was pretty close to an 
example I've previously given of something that was obviously just a 
typo fix - changing "highway=pirmary" to "highway=primary" - that 
wouldn't require a discussion with the community first.  However, people 
did object, both at the lack of discussion and of the size of the 
resulting changesets.

OSM is a community, and we need to work together to create the best 
map.  That means that we need to listen to each other, and if something 
that seems obvious to us but we're asked to explain, we need to do 
that.  It's clear (from the reactions afterwards if for no other reason) 
that it would have been helpful to discuss 
https://www.openstreetmap.org/changeset/108430155 beforehand, on a 
channel that's open to everyone in OSM.  Also "Dear Frederik Ramm, So 
the community has to ask for your permission each time we press ctrl+F" 
is probably not the best way to start a good-faith 
listening-to-all-sides discussion about anything.

Separately to all of that, you've mentioned "the huge and growing amount 
of bad data that OSM has" and replied to Simon's "Again that is just 
hyperbole, there is no indication at all of that" with "The problem is 
big enough to attract your and many other people's attention while I'm 
fixing it. I also provided a Wiki link to what I mean, maybe you should 
have a look at it".

For the avoidance of doubt, that wiki link is 
https://wiki.openstreetmap.org/w/index.php?title=Special:WhatLinksHere/Template:PossibleSynonym&from=0 
which is just a list of other wiki pages such as 
https://wiki.openstreetmap.org/wiki/Key:highway with no actual content.  
There's nothing there that describes in qualitative or quantitative 
terms "the huge and growing amount of bad data that OSM has".  It'd be 
good to know what you see the problem as.

Best Regards,

Andy (SomeoneElse in OSM)







More information about the talk mailing list