[OSM-talk] [Tagging] - RFC - Motorway_link implies oneway=??

Alex Mauer hawke at hawkesnest.net
Thu Oct 2 19:35:17 BST 2008

Matthias Julius wrote:
> I don't.  I think it follows the "principle of least surprise" better
> if implied values don't change too much.  

Great, then we should leave this as-is (implied oneway=yes for 

>If a highway is not oneway
>by definition oneway=yes should not be implied.

You do realize that "implies" means "if no value is specified, this 
should be the value that is understood", right?  That the actual tags 
over-ride the implied value?

So "the implied value could be incorrect some of the time" is no reason 
not to have an implied value.  That's why it can be over-ridden by 
applying the appropriate tag.  Having the implication is about reducing 
the effort required for someone to tag a motorway, and making it so that 
consumers of OSM data can make assumptions about the data.

There are three options:
1. make no assumptions: This means every single motorway_link needs to 
have a oneway=yes or oneway=no (or oneway=-1).  A pain for taggers, and 
doesn't help makers of routing applications who still need to handle the 
case where there is no oneway tag.
2. Assume oneway=no.  This means that 95+% of motorway_links must be 
tagged.  Still a pain for taggers. Makers of routing apps can't safely 
make this assumption anyway for fear of routing someone the wrong way up 
the motorway_link. It does mean that motorway_link has the same default 
as the other highways except for motorway though.
3. Assume oneway=yes.  This means only 5-% of motorway_links must be 
tagged.  It's the safe assumption that routers will have to make anyway. 
   And of course it doesn't prevent anyone who wants to from explicitly 
specifying a oneway key/value for every motorway_link if they feel like 

Looking at the above, assuming oneway=yes seems the only way to go.  Am 
I missing some big problem with it?

>  Trunk roads are
> probably mostly oneway, too ...

Off-topic.  This is about motorway_link.

-Alex Mauer "hawke"

More information about the talk mailing list