[Talk-GB] West Mids Ward Boundaries

[Andy Robinson (blackadder-lists)]

Ed Loach wrote:
>Sent: 07 April 2010 11:01 AM
>To: 'Steve Doerr'; talk-gb at openstreetmap.org
>Subject: Re: [Talk-GB] West Mids Ward Boundaries
>
>Steve wrote:
>
>> I'd prefer to see boundaries handled by an automated (and
>nationwide)
>> import process, as the OS data is likely to be at least as good as
>
>> contributors' own efforts and usually much better AND they are
>committed
>> to maintaining it going forward. Which seems a good opportunity to
>mention
>> that those of you working on the technical aspects of using the OS
>data
>> need to have a thought to how future updates will be handled. It's
>much
>> easier to insert a load of brand new data than to update existing
>data
>> from a refreshed set of source data in the future. You will have
>to be
>> able to distinguish new data to be inserted from existing data
>that may
>> have changed and existing data that may need deleting. Just a
>thought.
>
>As I suspect there won't be any ward data currently in, the idea
>sounds somewhat appealing, BUT I have not yet seen anywhere where
>imported boundaries have been done well and have spent many an hour
>tidying up the results (recently in Haiti and ongoing in GA, USA).
>Examples include adjacent areas using overlapping ways rather than a
>single shared way, an admin boundary being imported as a single way
>which then needs splitting (and any duplicate ways de-duplicating)
>before a relation can be created, and so on. If such an import were
>to be done, it should be done well or not at all.
>
Its also worth noting how the OS maintains this boundary-line data. Its not
apparently, as you might have thought, tied to OS MasterMap, which you would
think would provide precise positioning based upon land ownership etc.
Instead it's based off the 1:10k mapping, basically StreetView though not
necessarily that style. So importing boundaries isn't necessarily a good
idea and in any case is likely to conflict to some degree with existing
data. That's why prefer a manual process. There are not that many boundaries
to sort, so it's not like it would take years to do anyway.

The NaPTAN import has proved to me that we should never trust imported data
without our own verification.

Cheers

Andy