[Tagging] Points instead of areas

[Daniel Koć]

W dniu 07.08.2018 o 15:24, Christoph Hormann pisze:
> I think you have not understood the difference between measurement 
> tolerance and convergence here.

I'm not sure what do you mean by "convergence", but there's no
measurement tolerance problem, because without accepting area as a base
and without choosing centroid algorithm you don't even know what are you
measuring, in the first place.

If you followed next posts in this thread, you could see that the
"center" points are really not clear. I envisioned both choosing "right"
criteria and "right" spot problems, but I didn't know it's so well
illustrated already - Berlin center discussion shows it nicely, but Rome
shows exactly that your assumption of "social proof" fails hard in reality:

https://www.openstreetmap.org/user/dieterdreist/diary/40727

I like real life detailed data, but even in theory it's clear for me
that while people might want some points, they are just what I've said -
generalizations. For example in my city (Warsaw, Poland) we have
touristic center (Old Town), transportation center (Central Station),
probably biggest square near it (Plac Defilad) and historically highest
building there (Pałac Kultury i Nauki)... Which one is right? Well,
different one for different purposes. For label placing square seems to
be winner, but that's just one useful purpose.

So - no, point is not better than area and even not easier to verify.
Area seems to be more neutral for me - and here is where we can start
talking about measurement tolerance. For example area like 190 or 210 is
closer to 200 than 0 and the problem stops only when we zoom out a lot,
so it's all about 0. But this is generalization on bigger scale again...

> In any case how geographical objects that can be verifiably mapped in 
> principle are best represented in the OSM database is a matter of what 
> is most efficient, most convenient and least prone to errors *for the 
> mapper* (and not the data user!) to document the verifiable information 
> available about the object in question.

Sure, tagging should not be a royal pain, but I really don't understand
why you always praise the mappers at the expense of data consumers, as
if they were separated by wall and blessed/cursed, respectively. I also
believe such lack of balance would be simply bad for the project.

> For a plain rectangular footprint bench (>99 percent of benches) there 
> is absolutely no difference in the achievable level of accuracy and 
> detail in the representation of reality between these three variants.  
> But there is an immense level of difference in efficiency and 
> convenience of these representations for the mapper.

For me as a mapper convenience means to record the outline, not
measuring angle, width etc. For me as a map style coder it's the same...
But if I would like to make statistics about bench angles, it's different.

Something is wrong with this example, if I understand what point you
wanted to prove: the wall is not between producers/consumers and your
proposition is also harder for me as a mapper.

> Any argument beyond that (like that there is something inherent about 
> benches (or populated places or continents) that makes them more 
> suitable to be represented as X in the database) is usually just stuff 
> made up to convince mappers to map in a certain way for the convenience 
> of certain data users.

You speak only about most trivial case from my list. For anything else
the outline is more correct - buildings are sometimes rectangles, but
many of them have more complex shape where making a node and adding
measures would be not possible. Ditto for cities and countries, even
pitches have different shapes.

-- 
"My method is uncertain/ It's a mess but it's working" [F. Apple]