[OSM-talk] NoName

[Dave Stubbs]

On Mon, Sep 15, 2008 at 1:52 PM, Gervase Markham <gerv-gmane at gerv.net> wrote:
> Frederik Ramm wrote:
>> The absence of a name property from an object is not generally
>> significant (how many of our objects have a name - probably less than
>> 5%). Those that don't have a name simply don't have a name and it works
>> well for editors and renderers alike; nobody has ever even thought of
>> issuing all of them with a special tag documenting the absence of a name.
>>
>> The absence of a name becomes only significant in the context of a
>> certain check we run on our data, which operates on the assumption that
>> roads of a certain kind usually bear a name.
>
> I agree that the absence of a name is only significant in certain
> contexts for certain types of object - primarily roads. But I don't
> think it's just about "tests" - is the noname map a "test"?

Yes.

It's a very simple metric designed to highlight potentially unmapped
(just traced) areas. We have maplint, but I wanted something much
clearer to help direct mapping.


>
> The fact that this is mostly only a problem for names, and mostly only a
> problem for roads, makes me think that we have to look carefully at the
> trade-offs between a specific solution such as "noname=yes" and a more
> complex, general solution such as "test_ignore=name_present" which, you
> will have to admit, is not as obvious a tagging scheme!

There is an issue with noname=yes that it's "tagging for renderers" --
as with all such tags I don't mind them if it's really obvious that
that is what it is, and noname doesn't really manage that.
Just adapting Frederick's excellent observation slightly, maybe this
is more obvious:

validate_name=no

or something along those lines.