[Tagging] Micro- and macromapping with area=*

[Daniel Koć]

W dniu 31.03.2015 10:18, Martin Koppenhoefer napisał(a):

> I have for long been promoting a clearer approach for this kind of
> mess, and I agree that there would be need to implement some changes.

While my propositions are much more than this (I'd like to fix more 
fundamental issues), I'm happy to make at least a bit of cleaning. In 
the meantime I will try to convince you there's a real reason for bigger 
changes and we can gain something important that way (more data and of 
better quality at the same time).

I guess we deserve serious rethinking after 10 years, but if OSM 
community does not like it - just a few, rather critical responses 
indicates it - it's still very good to redesign some parts of the 
project.

> no, if you look above, "grassland" is a geographic feature, a
> landscape type, so belongs to natural, while landcover would be
> something like "grass", "bushes", "trees".

I was trying to see the mapping through the eyes of someone not that 
experienced like you and me and start from what he can see, not what we 
know he should have seen. I realize we have friendly editors like iD and 
JOSM with presets for everyday tasks, but they also do not help you 
decide in such cases. With so many choices, decision making process is 
hard for not advanced mapper and simply annoying for advanced one.

> if you speak English you could translate some pages into your
> language, although this is another big issue: translations tend to not
> follow up with amendments of the English version.

I ask myself a simple question: why we have this "small issue" at all? 
Why people don't translate even basic tagging schemes to all the 
languages that can be? If it's so easy, why the problem is so huge? The 
hidden obstacle must be really big if it outweighs this "easyness".

> yes, the world is complex, a grassy area could be a football pitch, a
> park, a meadow, a garden, a roof, etc., so you will have to understand
> and interpret and classify what you are mapping, there is no
> alternative.

TINA (there is no alternative) is another red-alert pattern. =} In my 
vocabulary it means: "Help, we're unable to leave the box we're in, even 
for a while!" ;-} Yes, there is the alternative and it's not trying to 
classify deeper than you really know.

The problem is simple: the more we push the people, the worse/less data 
we get - and we'll be not (read: "we already aren't") even aware of 
this! I'd rather have area=grass with no hints of grassy area status 
(which I would like to have) than nice and shiny landuse=grass, if that 
represents (lack of) the knowledge of a mapper better.

If we say: "sorry, bro/sis, life is hard, you HAVE to know!" she will 
abstain from adding some simple data (because we gave her no safe 
values, like in building=yes) or she will cheat the system picking more 
or less random/popular value. I consider both scenarios our systematic 
fail.

If someone is not really sure, I prefer to get the best he could give us 
- no more, but also no less. Someone else may come, see there's a 
general category and add more details later. But if we force him to 
choose no matter what, we'll have no data we could have or a bad quality 
data (lot of guessing inside) looking like a proper classification. We 
loose and we don't even know it! Trying to be perfect from day 1 
backfires exactly like this.

>> Again - I see some area populated with trees. It may be landcover
>> or landuse, but surely it is the area.
> 
> it will always be "landcover" AND "landuse".

So we have a simple hierarchy:

landuse (implies)=> landcover

but you have to know there is a "use" at all to tag it as "landuse". 
Since the trees may not be used (like in natural=wood), your statement 
is false.

-- 
Piaseczno Miasto Wąskotorowe