On Tue, May 13, 2008 at 12:57 AM, Andy Robinson (blackadder-lists)
<[EMAIL PROTECTED]> wrote:

>  Some of us map out an area completely in one go rather than doing it
>  piecemeal. Even if I come across some existing roads in a new area I ignore
>  them and do a new survey so that the whole area makes logical sence to me.
>  That way I can work out where the landuse areas are behind the houses and
>  the extent of school areas etc etc. So for me a reasonable level of
>  completeness is easy to decide and annotate.

That's how I work too. So when I mark the area that's complete on the
new system, I'll mark everywhere that I did as complete. That's a
no-brainer.

But let's scale this further, since what you, me and David have done,
while interesting, is still a small amount of what's there. I would
guess that only a proportion of the mappers will use this system,
let's say a similar proportion to those who map in
landuse=residential. But the key is that not everyone will use it.

So how do we scale the efforts of this subset of people who *do* want
to use the completeness system to provide measurements of the rest of
the data? How do you, me and David (say) work out which areas of
Glasgow and St. Louis are complete? Maybe we can look at the road
density and guess. But that could be automated. My original point was
that we can look at areas that we don't know particularly well and
it's much easier to spot the problems than confirm which bits are
done.

With the proposed system, it'll take me 15 minutes to mark everywhere
I thoroughly mapped, and then I want to do something useful. So I can
mark 20-something square kilometres of London as complete (to the
"95th percentile of completeness"), and *much more* as definitely
incomplete (i.e. Dave's renderings of unnamed roads) and much of it
'unassessed' where it could be anything from the 60th to 95th
percentile.

Anyway, that's just my take on it.

Cheers,
Andy

_______________________________________________
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk

Reply via email to