[OSM-talk] Superways again
john at mckerrell.net
Fri Mar 16 19:05:35 GMT 2007
I was putting together a response about how I thought perhaps it
would be better to just name these "groups" rather than "superways"
as I believe "superway" implies extra meaning that isn't intended.
Basically all we're trying to do is assign tags to groups of ways
without having to do it to each way, or have an easy method of
selecting multiple ways.
Then I realised that surely we have a method, tagging.
Purely as a hypothetical example which I'm definitely not
recommending, we could have superways by adding a superway tag to
ways. It would contain a number.
This assumes that we have facilities for selecting ways in editors,
or downloading way information, filtered by specific tags.
Once I realised that we already have the ability, I kinda realised
that we don't actually need this superway tag, we just need to
properly tag our information. For instance, want to select all of the
ways that constitute the M1 motorway? Do a query (API or editor,
whatever) for ref=M1 and highway=motorway. I realise that this would
pull out a motorway in another country that had the ref M1, but I
think that either that's valid, or the UK motorway should have
further tags identifying it as being UK specific which you could
further use. I don't know that an anonymous "group" would be a better
As an illustration, I'll go through your examples below showing how
tagging could solve this, but just in case anyone stops reading here
I'm not saying that this is is my final thought on how this should be
handled, more "I've sat down, thought this through myself and these
are my personal conclusions". Feel free to try to persuade me
On 16 Mar 2007, at 16:31, David Earl wrote:
> There are then lots of advantages to grouping such ways.
> * If you break a Way at a bridge, you can group the three Ways
> (either side
> and the bridge itself) so their commonality (it's the same road)
> can be
> represented (and so on along the road).
highway=unclassified and name="Seddon Road" and postal_code=L19
> * An estate road with many branches can be represented as a whole.
highway=residential and name="Rabbit Warren" and postal_code=E17
> * If you want to represent a bus route, the route tag in theory
> allows this,
> but in practice you can't put more than one route on the same Way,
> so it
> can't also be a different bus route or a cycle network route, and
> the route
> isn't coherent in any useful way - you have to search for where it
> next. So grouping ways to represent the concept of route would be
> Note this means ways can belong to more than one superway.
Hmm... I'm certainly less sure about this. Bus routes and cycle
routes should probably be separate tags, but multiple bus routes...
I'd be tempted to say this is a problem with tags not allowing
multiple values that we might look at solving.
> * A non-roundabout and non-node junction could be represented
> (together with
> its name or number) as a bag of ways (consider a grade separated
> carriageway with four slip roads, or a cloverleaf).
Give everything junction_ref=5 and ref=M6
Actually, another reason that I've just thought that creating
anonymous groups doesn't solve anything is that they're just that,
anonymous, how do we then label them? We add tags to them, but what
if someone creates a group with ref tag "M6", and then someone else
creates another group with ref tag "M6" in another country, now
they'll both come up.
Also, I realise that this all assumes we have ways to filter by tags,
but I think adding that capability to the API would be a lot easier
than bringing in a new data type. Also searching by tags can easily
be added to the editors (and of course is already there in josm,
though it could probably do with being extended).
Oh well, if David didn't like the blue touch paper I'm sure I just
have, I'm popping out now so I'll be interested to see if my mailbox
has blown up by the time I come back ;-)
More information about the talk