[OSM-talk] xybot

[Someoneelse]

From: Frank Sautter <openstreetmap at sautter.com>

> hallo someoneelse,
> 
> i'm the one behind xybot.
> 

Thanks - hi.

(re landuse/natural=wood/forest):

> you are right, i will take out those rules where xybot is just guessing
> the optimal values and ask the maintainers of
> http://tools.geofabrik.de/osmi/ or
> http://keepright.ipax.at/report_map.php to add them to their validation
> tools so a human brain can set them to the correct value.

Thanks.

> 
>> Another problem is where there is:
>> tag_a=value_a
>> tag_b=value_b
>> If it thinks that tag_a is a misspelling for tag_b it'll change the 
>> tagging to the following:
>> tag_b=value_a
>> which is just plain wrong.
> no, such changes are prohibited in the algorithm.
> 

The reason that I mentioned that was due to changes to way 16359613. 
xybot did highlight an error that I made here (which is good) but 
unfortunately resolved it the "wrong" way.

Here a river (the orginal way 16359613) flows northeast into an area 
which is a man-made landscape - it's drained by pumping water out along 
a series of canals referred to locally as "drains".  A later section of 
that river (some distance away) had been mapped as "water=drain".  While 
you could argue that the "water=" part is wrong, the "=drain" part is 
certainly correct for that part that had been mapped, as would "=river" 
be - it's both.

I added on some of the intermediate bits from NPE and joined the two 
sections.  This resulted in "waterway=river, water=drain".  Whilst 
accidental, this does correctly give the sense of how that stretch of 
water fits into the landscape.  My error was to let that apply to the 
"natural" part of the river as well as the artificial part (and it may 
need someone closer to it on the ground to work out where that is), and 
to let the source tag that I added propagate too far west.

What seems to have happened is that on April 8th way 16359613 was 
changed from:

waterway=river
water=drain

to just

waterway=drain

I've since changed it to waterway=river, which is more correct but 
doesn't communicate the "draininess" of the northeastern section.  Next 
time I'm up there I'll have a go at improving things, including adding 
some of the many other rivers, canals and drains that aren't currently 
fully mapped.

>> If it is necessary to "validate" the data against a pre-conceived idea 
>> of what it should contain (and I'm not convinced that it is) wouldn't it 
>> be better to collect lists of "errors" by last modifying user and 
>> suggest via mail that they review them? 
> i think this would result in a massive spamming of the osm users and i
> don't think that the messaging infrastructure of osm is able to handle that.
> 

I don't think that telling someone that you're going to change something 
that they've mapped could be construed as "spamming".  If you're making 
so many changes that OSM's messaging infrastructure would be unable to 
cope maybe you shouldn't be making so many changes?

You could argue that trimming a space off the end of a name is always 
valid, and maybe certain typos (although people have argued against that 
on this list in the past), but beyond that there really needs to be 
communication.  At the very least, this would enable discussion of, for 
example, why "notes=" isn't valid but "note=" is - if people don't know 
that "incorrect" tags have been changed they won't use the "correct" 
ones next time.

The tags used in OSM should be used by consensus.  If someone wants to 
use a tag, it's valid.  If more people use it, it's more valid.  If 
someone decides to start mapping something as X instead of Y because 
everyone else maps it as X, that's a good outcome - but for that to 
happen they've got to know about it and not have stuff "randomly" change 
on them behind the scenes.

Cheers,
Andy