[Talk-GB] Guidance for adding PRoW to OSM: prow_ref=

[Robert Whittaker (OSM lists)]

On 1 January 2013 16:30, Craig Loftus <craigloftus at gmail.com> wrote:
>> 1/ prow:ref suggests some sort of name-spacing, but we haven't
>> actually developed any tagging scheme that makes use of a prow:*
>> name-space. So currently prow:ref  would be the only tag used.
>
> Is it wise to preclude adding more tags to the namespace? As an example, one
> additional tag that occurs to me is "prow:operator" (or "prow:authority"),
> to describe the local authority the references 'belong' to.

I wouldn't have thought that listing the authority would be that
useful -- you should be able to work that out from the county that the
way resides in. Apart from something like prow:type (for which we
already have the established designation=* tag) nothing else springs
to mind as being Rights of Way specific. If anything else is found, I
don't see a problem in having a later proposal to introduce a set of
prow:* tags and in the process change from prow_ref to prow:ref. So I
don't think it's necessary to use prow:ref "just in case" at this
stage.

> I agree source:prow:ref looks ugly, but I am not clear what is ambiguous
> about it?

Is it the source for prow:ref or is it a ref value somehow relating to
a source:prow namespace? Granted this particular tag is probably not
likely to be mis-interpreted, so this is only a very weak reason.
Ugliness was my main concern here. (The subtle issue is using using :
for both namespaces and recording sources, which have slightly
different semantics, but it's too late to do anything about this in
OSM now I fear.)

>> 3/ prow_ref mirrors other ref types in use, such as bridge_ref,
>> route_ref, ncn_ref, and local_ref, which are generally used rather
>> than the alternative colon separated versions.
>
> This seems like an appeal to popularity; one could point to tree:ref or some
> other *:ref.

There's a difference between appealing to popularity on a
non-established tag where numbers are likely to be decided by a small
number of mappers who happen to have chosen one over another for a
variety of reasons, some of which may just be copying any other
instance they found. As opposed to looking at well-established tags
(and patterns of tags) which are widely used and would now be very
difficult to change.

If you look at the numbers of uses in taginfo, you'll see that *_ref
is much more widely used than *:ref. For example, there are only four
*:ref keys with over 10k instances, and two of them are source:ref (or
a derivative thereof), which is arguably different. There are 15
different keys for *_ref with over 10k uses.

Anyway, that's more of an explanation of why I think prow_ref would be
preferable. If other uses are found for a prow namespace I might be
convinced to change my mind. We do need to settle on one tag to use
though, and I'll be happy to go with whatever consensus emerges. I'd
suggest we ask the others who have been making use of either prow_ref
or prow:ref (or even just ref=*, as I did originally) on Rights of Way
for their opinions and their reasons for choosing the one they did.

Robert.

-- 
Robert Whittaker