[Tagging] Boundary Relations. What's a subarea used for?

André Pirard A.Pirard.Papou at gmail.com
Sun Jan 11 20:42:04 UTC 2015


Hi,

Look at the Belgium relation
<http://www.openstreetmap.org/relation/52411> and, while hiding the
subareas (in left pane), try to figure with that map the *administrative
tree* (regions, provinces) using the borderlines.  You won't.
Now look at the subareas and something you will notice is that Brussels
is not inside any area. It has got its own independent status. Not
obvious at all from borderlines.
Now, trying to figure the tree, you can use your knowledge of Belgium or
try to right-click on each subarea to open its link in another browser
tab and you will find that Belgium is divided in two regions: Flanders
in the North and Wallonia in the South.
Now what about the other subareas?  Try to jigsaw puzzle them again and
you will fit Flemish Community in the North, French Community in the
South and German community to the East. And so what about Brussels
again? It's Flemish-and-French bilingual.
So,  Belgium=Flanders+Brussels+Wallonia regions.
and Belgium=(Flemish+bilingual+French+German) communities
Similarly, United Kingdom=England+Wales+Scotland.
I may miss a few UK details, but that's because ... it has no subareas
to see.
And I recall a guy from Munich.  Wasn't his problem that Munich is not
part of its Land?
Isn't the Munich problem like Brussels (please do not focus the thread
on that if it's wrong).

And the same process can be repeated at each boundary level down the tree.

Moreover, it is straightforward for a consumer program like OSM.org to
use the subareas to draw the outline of the regions, provinces, etc.
inside the country map or to do other things like measuring borderlines.
Why the heck would we remove the subareas and cause those programs to
stop working, instead of adding the subareas which is really
straightforward too?

Now, there is a very interesting property of subareas making a nice program.
Choose a boundary relation. Pick its subareas members one by one and put
their borderline ways (members) in a bag.  Now, remove those ways that
appear twice in the bag. What have you got?  The borderline of the
chosen relation.
For example, by combining that way the borderlines of the UK subareas
England, Wales and Scotland, you get the borderline of United Kingdom.
Idem down at any level.
So, it looks like masochism somehow to tag borderlines for anything else
than the lowest level. From it, one may have that program compute the
borderlines of every relations upwards.
But I guess software having to build a country boundary that way would
have a hard CPU time.
The idea that would come to mind is to keep the borderlines as a cache. 
But we have the cache already made manually. Why not simply let the
mappers use the program to build the borderlines manually? They don't 
change often.

That program can work two ways.  Either to compute borderlines. Either
to check that the intricate borderlines match the so easily tagged,
error proof subareas.
There is presently no real QA program for borderlines. Here it is. Ad
it's soooo easy...
Start checking Belgium and France, Osmose!

It's a really simple program, a very few tens of lines. I didn't write
exactly that but that was close (computing total length of boundaries in
a province or country). But, shame on me, I misplaced the source :-(

I once read on some server a French text wondering whether the right
solution is subareas *or* borderlines.  As it often happens, the answer
is hard to find because the question is wrong. It is not "or" but
"*and/or*". And the answer is "and", both.
Also wondering which makes tagging boundaries the easiest.  Certainly
creating the boundary relations with just simple subareas linking them.
Then start adding borderlines at the lowest level.  And run the check
program above working at progressively higher levels.  That programs
detects incomplete borderlines loops and hence dangling lines near which
more have to be filled in. Visually checking with the spoken above
rendition that there are no unwanted holes also checks the subareas tree.
That's how we, Marc, André and Patrick (yes, MAP!) made the boundaries
of South Belgium.

One thing is certain.  Should anyone remove our subareas because, as
said, he does not understand, it would be like stealing a worker's tool
and I would stop any boundary work immediately.

I won't write too much in one article but I'll add this.  What's that
aversion against redundancy?
Redundancy used as crosscheck is used in many place. For example, TCP,
which is the transport protocol of the Internet, uses redundancy to make
sure that these words I wrote came to you intact.  Same on a disk drive
surface to make sure the recording is correct.  Etc.
Should we remove the non-English pages of our wiki because they are
redundant?
No, because they have different usage and it's the same with subareas
and borderlines.
Except that for the latter we have a program to check consistency.

Cheers

André.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20150111/00e3b43a/attachment-0001.html>


More information about the Tagging mailing list