[Talk-us] Nominatim in CDP

[stevea]

>CDPs in OSM have been an ongoing issue of discussion for a while.

Yup, I wrote here about CDPs getting tangled up with admin_boundary 
in OSM in 2012!  CDPs have no effective or "OSM sensible" admin_level 
boundary value, I think that much we have established.

>NE2 stated that he would delete them all unless someone could show him
>a single example of them being useful.

Cleaning up messes in OSM (whether by NE2 or not) is something I'm 
familiar with.  Resulting semantics (behavior of how OSM consumers 
use data in it) can be quite complex:  in some cases CDP data 
confuse, in some cases CDP data might be "all that there is" to get a 
fix on "where is this?" in a geocoding environment.  Usually, going 
wider (to a county or a state level) might be more strictly accurate, 
but without the granularity desired.  For example, you might be able 
to "get correct" an admin_level=6 but, 7, 8, 9 or 10, no.  (CDPs 
might be thought of as "somewhere around 7, 8 or 9" but that becomes 
nonsensical when you truly lean hard on it).  There really are a 
variety of ways OSM's data can be queried and parsed to make sense of 
things, and CDPs can both muddy the water and occasionally provide 
"something" which MIGHT BE better than "wider (but correct) something 
else."  With the data as they are today, you just don't always know 
that in advance.

>I pointed out that Bethesda, MD (noted for being where the NIH and the
>Naval Medical Academy, along with several other large landmarks) is a
>CDP, as is Silver Spring, MD, which is the 2nd most dense place in MD
>except for Baltimore.
>
>After some discussion, he agreed not to delete the objects in the whole US.
>
>The general feeling from many people were that the CDPs were useless
>information- only interesting to the census workers and not the
>regular people on the street. For them, it probably made sense to
>delete the CDPs. For places where CDPs do make sense to keep in, it
>would be sad if someone deleted them, but that's likely what happened.

See:  there IS an odd dichotomy of cases where "using CDPs make some 
sense" and "CDPs don't make sense."  Usually, the latter, but on 
occasion, more than once, certainly, the former holds true.

>This is (yet another) reason why I believe so strongly in Ian's effort
>to move government boundary data out of OSM and into another dataset.

As tools to move brief (often, relatively brief) snippets of 
now-in-OSM data get better, easier to use (for intermediate and even 
novice mappers) and better community consensus hones in on which 
datasets do and don't belong in OSM, I think we shall see more and 
more migrations of the sort that Serge refers to.  There is a "bare 
bones" of data which truly do belong in OSM.  There are also data 
which do not belong in OSM.  Harmony about which and where can be 
difficult, as we have seen, but these efforts can and do move us 
forward.  Serge sees a future (of more and more 
not-in-the-OSM-database data), and it is here now.

What are good ways to keep this happening?  Good communication and 
deeper and wider understandings about how we use our data.  And 
thanks to those who work on OSM data software tools (editors, style 
sheet manipulators...), promulgate them and document how to use them. 
Keep up the good work, everybody!

SteveA
California