[OSM-talk] Worldwide non-surveyed tag edits

[Peter Wendorff]

Hi,
I think two issues are mixed here that shouldn't be.

One thing is the question about mechanical edits, the other thing is
more about unifying edits.

The latter occurs most often on mechanical edits, but it seems the two
issues are seen as equal what I don't think they are.

I would combine both to "algorithmic" edits.
It's not different wether I use overpass, josm selection by filters and
beyond that mouse and keyboard to change all tags from A to B or to
delete objects that correspond to a certain condition or if I do that
using a fully automatted script.
In both cases it's dangerous to make a mistake. And in neither case I
can or do any individual check for each object.

In both cases - but, yes, even more on the bot/script way of doing that
stuff, is the danger to make errors.
The first error class is a technical one, which is more likely on the
script approach then on the nearly-manual mass editing.

The other class of errors occurs earlier: When deciding to make an edit.

Changing amenity=restaurante to amenity=restaurant (no idea if that's a
common real-world bug) seems to be clearly a good way to achieve, and
there are typos that are definitivly typos. But where does a typo get a
semantical difference?

Sometimes even people unify stuff by look-alike semantical identity,
which, looking more closely on the issue isn't. In these cases we have a
problem in the tagging perhaps, but not at first a problem of synonymous
tags, but about IF they have the same meaning.

To conclude:
There are IMHO at least two reasons why large scoped edits should be
discussed and therefore be in some kind of policy:
If you fail on a large edit, you break a large amount of data.
That's why you should let others look and verify what you're going to
1) make sure no technical bug appears
2) make sure you didn't oversee something, e.g. a different meaning of a
tag or tag combination.

regards
Peter

Am 11.06.2014 15:06, schrieb Jochen Topf:
> On Mi, Jun 11, 2014 at 01:55:41 +0200, Christoph Hormann wrote:
>> On Wednesday 11 June 2014, Jochen Topf wrote:
>>>
>>> I think the "mechanical edits policy" has stepped over the line here.
>>> A mechanical edit is one where somebody uses a special program that,
>>> based on some simple criteria, does *automatic* changes. Using
>>> existing tools like JOSM and XAPI to find problems, looking at them
>>> manually and doing edits, is not a mechanical edit and should not
>>> fall under that policy.
>>
>> And i think it does not.
>>
>> The policy is probably not worded in the clearest possible way but it 
>> has always been my interpretation that the key question is if the 
>> modifications made to the database are individually verified by the 
>> mapper, not if you use some fancy filtering to find those objects you 
>> want to modify.
> 
> I think we are probably in agreement here. My "looking at them manually" might
> be a very bad wording, but it meant the same as your much better "individually
> verified".
> 
> But the policy first disallows all of this and then makes the exception "if you
> check each individual action caused by this" then its okay. Thats the wrong way
> around. It means I have to defend myself when doing this. It means I have to
> prove that I am not doing anything wrong. Of course I have to verify that what
> I am doing is okay, that is the "default" anyway. But we are still measuring
> with two different sticks if we single out edits based on filters or whatever
> and not treat them like edits based on 20 year old sat images or whatever.
> 
> Jochen
>