[Tagging] Deprecation of associatedStreet-relations

[moltonel 3x Combo]

On 24/01/2015, Tobias Knerr <osm at tobias-knerr.de> wrote:
> On 24.01.2015 13:12, moltonel 3x Combo wrote:
>> Recomended isn't mandatory. The name tag of associatedStreet is only
>> of use to mappers (to find the relation in the editor), not consumers.
>
> Not mandatory, but still used in 93.20% of relations.

Yes.

> So ignoring the relation is not practically feasible.

I don't get that ? Ignoring the relation *name* is not only feasible,
but the right thing to do for consumers like nominatim.

>> "hindered" is a matter of perspective. With associatedStreet one can
>> change a house's street address using only the mouse. When a house is
>> already in a relation, changing its housenumber is just as easy as
>> with addr:street.
>
> It's not a matter of perspective that the most common use case (adding
> addresses from scratch) is still not helped by relations. Clicking on
> the building and entering number and street name in a form is also a
> very natural interface for those not familiar with advanced tools,
> making any overhead pointless.

You see "hindered", "not helped", "overhead" where I see a workflow
that is both simpler (less typing, which is nice on desktop and
invaluable on mobile devices) and more powerful (preparing a
housenumber survey by armchair-mapping buildings and
associatedStreet). I know full well that many people can't place
"simpler" and "relations" in the same sentence, but it's not a
universal fact : it is a matter of perspective.

>> Basic tasks are never going to be relation-free. Multipolygons, bus
>> routes, turn restrictions... The more complete osm will get, the more
>> impossible it'll be to avoid relations.
>
> Even if all bus routes and turn restrictions were mapped in OSM, they
> would still only affect a small amount of roads in the database. Mapping
> every address with associatedStreet relations, on the other hand, would
> affect almost every street in a settlement.
>
> In fact, if every address today was mapped with associatedStreet
> relations, that relation type would be more common than every other
> relation type combined!
>
> So avoiding associatedStreet would of course not remove the problem
> entirely. But it would make the problem so much smaller that getting rid
> of it is absolutely worth it.

I'm not sure which simile will make most sense to you, but in the
domain of databases and programming it is considered very important to
detect errors as soon as possible, so that they can be corrected
before they spread to other data or the related information goes
missing.

Here we have "newbies have trouble with relations" as a source of
errors. If relations are rare in the db, it's be a long while before
the newbie notices, and he'll have messed up plenty of data. If they
are common, mappers will realise earlyer that they need to be taken
into account, and will commit less errors overall. If they are common,
tool support will improve and the number of errors will diminish.

Removing associatedStreet from the db today will only remove 6% of
relations. That's hardly going to make a dent in the error rate.
Improving editor support (I gave iD a try yesterday and was surprised
at how cumbersome it is for relation) should yield much better
results.