[OSM-talk] Proposed feature for noname

[Christoph Böhme]

Hi,

thanks for putting this proposal together, Rory!

So far I followed the whole noname discussion quietly, but I am getting
the feeling that nobody is advocating what I think is the only clean
solution of the problem: Frederik's "ignore_test" suggestion.

The other solutions made so far essentially suggest to add a negative
image of the world to the database: Not only saying what is there but
also what is not. Consequently this would mean to tag streets not
only with the features they have but also with oneway:absent="yes",
cyclelane:absent="yes" and so on to indicate that someone has checked
that a feature is definitely not there. Take this to the very extreme
and we would even have to add points saying postbox:absent="yes" to show
that we checked that there is no postbox in a location.

All this information about absent features is completely irrelevant for
users of the osm data (or of maps in general). They generally assume
that the absence of information on a map indicates that the feature
does not exist. The fact that absent information could also just mean
"we have not checked yet" would not help when someone want to use a map.
Just try to find an application for a map that tells you that it does
not what the world looks like. This only helps you to find out when
better not to use this map.

There are much better ways to rate the quality of a map than just
collecting data about the correctness of each and every object on 
your map (which is quite error-prone itself!). Based on observations
how features normally accumulate in the real world we can devise tests
and see if our map differs from this normal distribution of features.
So e.g., instead of collecting information about certified postbox-free
areas, we could just find out how many postboxes are normally on square
mile and than compare this with the distribution of postboxes on our
map. The only problem with this approach is that there might be outliers
where the test tells us that there are too few postboxes, but we
checked and there are actually no more postboxes. In this case we want
to tell our fellow mappers that the test does not work for this
particular square mile and save them the time of checking it over and
over again. 

So, essentially we do not need tags to describe the absence of
features, but only tags that tell us that it is okay for test XY to
fail in this specific case (Making the postbox test above suceeding
without adding dummy (non-)postboxes, would be difficult. This I
would call tagging for tests). Therefore, I strongly argue for using
Frederik's solution of tagging streets that have no name with something
like test_ignore="street_has_name".

My main concern is that we should not start rating the correctness of
each and every tag in the database as a measure of quality assurance,
but rather use statistical tests for this, which are a much better way
to assure the quality of the osm data. 

Christoph


Rory McCann <rory at technomancy.org> schrieb:

> Hi,
> 
> What to do about a road that doesn't have a name comes up on the list 
> frequently. There's a few options suggested.
> 
> I've created a new proposed feature for this listing the available 
> options (noname=yes, name:absent=yes, missing_tags=name, etc etc)
> 
> http://wiki.openstreetmap.org/index.php/Proposed_features/Noname
> 
> Please flesh this out and use the discussion page. If one is the
> clear favorite, it should be voted on.
> 
> Hope that helps
> 
> Rory
> 
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk