[Talk-GB] [OSM-talk] Improving ref=* documentation

[Colin Smale]

> On 08/05/2021 10:07 AM Mike Baggaley <mike at tvage.co.uk> wrote:
> 
>  
> >If you have a road which is signed (for example) "A12" for almost its
> entire length, but somewhere there is a one-off sign that says "A21", do we
> tag that bit of >road as "A21"? Over what length? Or do we map following our
> cognitive processes, and assume that the sign is erroneous?
> >
> >If you have a road that in fact used to be the B2009 but was declassified
> years ago, but somewhere along its length there is a rusty fingerpost in the
> hedge that >has the old number on it, does that road magically regain its
> number from 30 years ago?
> >
> >If we are not going to let many decades of data modelling experience get in
> the way of our tagging schema, we accept that there is only one "ref" for a
> road. How >we judge which one to choose is what we are discussing here. Most
> arguments seem to revolve around a use case whereby a car driver is
> navigating, looking at >signs to help decide which way to go. The human
> brain is good at glossing over mistakes that appear obvious, but that's no
> reason to propagate them.
> 
> I suggest using name:signed and ref:signed to hold incorrectly signed
> values. You can then have name:signed=yes, name:signed=no and
> name:signed=<signed name>. I would suggest that if used to hold a value
> other than yes/no, then source:name and/or source:ref ought to also be
> specified so that it is clear why the name/ref is not the same as the sign.

Problem solved, then. Thanks for the clear and pragmatic solution. ref and name carry the official, proper values, and if a sign disagrees, put that in name:signed or ref:signed together with information about the sources.

What's not to like?

This allows two different renderings to be derived from the same data - one using the official values, and one using the "as-signed" values. The consumer can choose, according to their specific use case. We provide the data for both.