[Talk-GB] England Cycling Data project: DfT cycling data now available for merging

[Andy Allan]

On 18 June 2012 14:10, Graham Stewart (GrahamS) <graham at dalmuti.net> wrote:
>
> smurph wrote
>> I've just been looking through the CUBA data and I think we need to show
>> that a route is part of a relation (specifically NCNs - which are mostly
>> done by relation in the Bristol area) to avoid someone retagging all of
>> the ways as NCN when they are already part of an NCN relation.
>
> Similar situation in Northumberland: the NCNs round here are all in
> relations.
>
> In fact the "Cycle" tab on Potlatch2 treats all cycle routes as relations,
> so it is likely to be very common across the country. Perhaps the tool could
> be modified to take account of the "network" and "ref" tags on any
> "type=route"+"route=bicycle" relation applied to the way? Or perhaps just
> warn against merging *_ref tags when a way also has type=route relations on
> it?

Yes, handling the route relations isn't ideal. It's something I've yet
to solve to my satisfaction, given that we have a dataset with route
information on tags, and widespread use of route relations in OSM
itself.

I think the best thing is for people to use common sense - you can
mark a DfT line as "complete" without having to have the tags
matching, so in the situation described, there's no need to add tags
to the way.

I've tried hard to make both the p2 work and the snapshot server
completely generic and re-usable, so I'd like to avoid any hard-coding
of tags into either component. At the moment there's no side-by-side
reconciliation for relations within p2, so that would be the first
thing to work on. After that, some sort of uber-magic reconciliation
of a way+tags vs a way+relations would be required, and that's going
to need some serious work!

> Will the errors/discrepancies we identify be fed back to the DfT?

Unless Martin knows more than I do, then in all honesty I doubt it.
I'm hoping instead that whenever the DfT next want cycling data - say
2-3 years down the line, they won't commission a separate dataset and
will instead use OSM directly.

Cheers,
Andy