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

Colin Smale colin.smale at xs4all.nl
Fri Aug 6 11:19:42 UTC 2021


> On 08/06/2021 12:28 PM Frederik Ramm <frederik at remote.org> wrote:
> 
>  
> Hi,
> 
> On 06.08.21 12:13, Colin Smale wrote:
> > Ahem... There are plenty of examples of (partially) subjective information in OSM. Tracktype and smoothness for example. Even highway=* gives rise to discussion from time to time, as one mapper's judgement differs from another mapper's.
> 
> Nonetheless we should always *strive* to avoid subjective judgement, and
> introducing tags that invite more subjective judgement is certainly
> something that would need very good reasons to go ahead.
> 
> > Only by reference to an authoritative source can all elements of personal judgement be eliminated from the equation.
> 
> Unsure what you understand as an authoritative source here. If this is
> referring to "what someone else has in their database" then no, such
> authoritative sources don't square with what we're doing in OSM and
> while they might eliminate personal judgement, they introduce other
> problems. (We want to map what *is*, not what someone else thinks or has
> recorded.)

What someone else has in their database *where they "own" the data*. So ref=* on a highway referring to a government source is fine by me because the government source owns the data - if they say it is the A123, it *IS* the A123 (assuming no errors). We can tag source:ref=government to assert that the ref=* value is correct *according to that source*. What I have got in my personal database on the other hand bears no weight as I have no authority to issue new road numbers. Maybe you are suggesting we no longer put ref=* etc. on stretches of highway, but just record the presence of physical signs?

Frederik, I hope you won't mind me saying that you tend to have a very puritanical view on this subject. IMHO that can only work if OSM contains a complete and correct model of the whole world, which (although it might be a noble goal) is never going to happen. In any other case OSM's data will frequently need to be blended with data from other sources, which is where "foreign keys" (i.e. ref=*) come in. Their sole purpose in life is to make that link, and they are by definition "what someone else has in their database". Some true attributes are potentially different, as their presence in OSM could be considered a denormalisation where the object is "owned" by an external party. However we all tolerate their inclusion in OSM for reasons of practical usability, and some pragmatism and nuance in our attitudes is essential.

> If you are saying that the authoritative source for the width of a road
> is a tape measure, then yes, that's an excellent authoritative source.

Yes, I mean an objective source that "cannot be argued with." What we normally call "verifiability." A random mapper would come to the same conclusion. If this random mapper is able to come to a different conclusion, then they are wrong, unless they are actually using a slightly different frame of reference, possibly without realising it themselves.



More information about the Talk-GB mailing list