[Talk-us] Time to retire ref= on ways?

[Apollinaris Schoell]

fully agree we should keep this target in mind.
But first we have to resolve a long list of problems first. 
there shouldn't be any time when the renderer or other data consumers will be left with completely broken data because step2 was done before step1
osm doesn't have any way of enforcing anything we need to be careful to kill the dinosaur too early 


1) route relation tagging has to be defined, agreed and accepted widely. currently it's a mess.
2) rendering, garmin maps, any other major data consumer must be updated to use relations. currently none does to my knowledge. no wonder since 1) isn't done
3) define a grace period after 1,2) is done and consider to delete them after that. No need to do it because any consumer understanding relations the right way will push down the relation ref and ignore the way ref.



On 8 Mar 2010, at 2:36 , Paul Johnson wrote:

> It's time to retire ref=* on highway=* ways to describe attributes
> of the overlying route instead of the physical attributes of the way
> itself.  Using the ref= tag on ways to describe routes simply
> creates more problems than it solves for many reasons.
> 
> * The ref=* tag on a way is describing properties of a route that
> is using the way, not a property of the way itself.
> 
> * Many bridges and tunnels have signed references that would
> actually be physical attributes of a way, but with the ref= tag on
> ways describing the overlying route instead of the way itself,
> makes it impossible to properly describe these attributes if ref= on
> a way is describing the route above the way, not the way itself.
> 
> * The ref= tag as defined for ways now includes more than the ref,
> but also the network.  ncn_ref, int_ref, etc were created as an
> attempt to describe network references uniquely, but there aren't
> *_ref keys for every possible network already in play.
> 
> * The US has two federal highway networks, each state has it's own
> highway network, and counties and cities have the option for their
> own local networks.  That's at minimum 52+ *_ref keys that would be
> needed to describe each network uniquely...for the US alone!  And
> we're not even into transit or other routes that might use the way!
> 
> * Munging the modifier=, network= and ref= tags provided by
> relations into a single do-all ref= tag creates more problems than
> it solves, particularly for formatting.  It also creates
> hard-to-answer questions for renderers and parsers.
> 
> * Multiple routes, particularly when they are involved in multiple
> networks, creates unmanageable way ref= tags. It also makes it
> more difficult to describe attributes that belong to the route,
> not the way itself (such as which direction it's going, whether it's
> a bypass, business, toll or other sort of route, etc).
> 
> Given that we have route relations, and have had them for some time
> now, perhaps now is the time to:
> 
> * put ref= information pertaining to the route that travels on the
> way to a relation for that route.  Provide facilities to search by
> network and ref on relations.
> 
> * Actively remove ref= tags describing routes from ways that have
> route relations already:  Let's kill this dinosaur.
> 
> Thoughts?
> 
> 
> 
> _______________________________________________
> Talk-us mailing list
> Talk-us at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-us