[OSM-dev] segment discussion - fork of mysql partitioning

Richard Fairhurst richard at systemeD.net
Thu Aug 31 09:09:42 BST 2006


Andreas wrote:

> i understand this, but segments are useful for sub-informations like
> bridge=yes, tunnel=yes, etc. which do not apply to the whole way - a
> possibility around this would be to make ways groupable (arrays of
> ways), which is technically not very different to having segments
> grouped..
>
> how would you handle these cases without segments? have a seperate way
> with same properties + the extra properties (i.e. like tunnel=yes)

Yes, absolutely. As Steve said in his early-August mail, this seems to  
be what a lot of users are doing anyway.

I'm not sure what grouping ways would lend to the party - surely that  
implies a primacy of one attribute over another? To take an example,  
the A55 in North Wales is dual carriageway, occasionally with one  
carriageway in tunnel, another not. You could, theoretically, group  
them all as "one grouped way". That makes sense in a route-planner,  
when it's immaterial whether you're going through a tunnel or not. It  
doesn't make sense when rendering a large-scale map, though, because  
you then want the tunnel to be shown differently (probably = = = =)  
from the overground road. Why would we want to design a data structure  
that makes sense in one use case but not another?

cheers
Richard





More information about the dev mailing list