[OSM-dev] Developers requested to help provide "completeness" tools

David Earl david at frankieandshadow.com
Mon May 12 12:55:33 BST 2008


On 12/05/2008 11:59, Andy Robinson (blackadder-lists) wrote:
> Posted to talk and dev
> 
> This post really follows on from my consideration of Completeness metrics. A
> number of experienced OSMers who have mapped out there areas have either
> contacted me directly or posted through the lists about areas of OSM that
> are essentially complete, at least as complete as you would expect from a
> conventional on-line map or better.
...

I've raised this a number of times in the past and have been thinking 
about it recently - to the point I was thinking about implementing 
something.

Here is what I was going to do:

1. use a set of ways like coastline (i.e. with an "on the right" rule, 
because they'd be too long as one way), to define areas of completeness. 
By definition, coastline would form one boundary.

2. Render these plus coastline to a new set of tiles (possibly only at 
one zoom level, say 13 or 14) so that complete areas are transparent and 
incomplete areas are semi transparent red, say. Use the ocean tiles 
database to avoid putting red over all areas of sea.

3. Present these tiles as a layer. For other zooms either combine or 
divide, or sample, or simply rescale in the browser - the edges need 
only be quite coarse.

I felt this leveraged most from the tools we already have so would be 
the most straightforward to implement.

I was going to have only one measure of "complete", that is the surveyor 
asserts that all publicly-accessible roads are present, with names for 
all except for those impossible to determine and when that is indicated, 
and all key points of interest from a limited set: schools, pubs, places 
of worship; and any railways and significant watercourses.

I think it would be confusing for a consumer to have lots of variations 
in what "complete" means - just an indicator to say "you can't really 
trust this area yet" would be simple.

However, since the boundaries can be tagged, there would not be any 
problem about expanding this for internal use to cover degrees of 
completeness.

David




More information about the dev mailing list