[Talk-GB] Admin Boundaries and OS OpenData BoundaryLine

[Chris Hill]

On 29/05/12 16:00, Tom Chance wrote:
>  Most boundaries follow existing features like roads, rivers, etc. 
Not always by any means. Many urban boundaries follow roads, but many 
rural ones run alongside roads and have little jinks in them where they 
cross to the other side of the road. This allows a stretch of road to be 
firmly the responsibility of one administration, not shared. Many 
boundaries follow an old course of a stream, when the stream moved the 
boundary did not.
> They need to be manually entered as relations sharing nodes with those 
> features. 
I would say that sharing nodes can lead to problems. Boundaries that get 
imported or manually traced from OS data often have no visible reference 
on the ground. If you share nodes with something else, when someone 
aligns that something else to aerial imagery, or a GPS trace or 
whatever, the boundary (which was probably right) gets moved too. Why do 
nodes of one object need to be shared when they are quite different 
objects?
> In my experience this is often a nice opportunity to spot other 
> problems with very old features using aerial imagery and GPS tracks, 
> e.g. poor alignment, or complicated junctions that aren't fully 
> modelled for routing. So much better done manually than by dumping a 
> load of new ways into the database.
I agree that manual scrutiny is vital. Local knowledge and control is 
also important. Documenting the existing practice would help too.
>
> 2) Many boundaries already exist, but are often slightly incorrect, 
> e.g. not sharing nodes with existing features but being a little offset.
That offset might be right. See above.

[...]

-- 
Cheers, Chris
user: chillly