[OSM-talk] superways as relations ?

[Peter Miller]

> Date: Mon, 18 Aug 2008 09:16:47 -0700
> From: "Karl Newman" <siliconfiend at gmail.com>
> Subject: Re: [OSM-talk] superways as relations ?
> To: "Robert (Jamie) Munro" <rjmunro at arjam.net>
> Cc: talk at openstreetmap.org
> Message-ID:
> 	<52ae42fa0808180916h63a872a6ude6c64e5f31f722f at mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
> 
> On Mon, Aug 18, 2008 at 9:00 AM, Robert (Jamie) Munro
> <rjmunro at arjam.net>wrote:
> 
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > David Earl wrote:
> > > On 04/08/2008 11:14, vegard wrote:
> > >> For naming of streets in cities, where properties change very often
> and
> > >> you have to make many small ways, it sometimes gets annoying that the
> > >> name is duplicated.
> > >>
> > >> I was wondering: How good/easy would it be to make a superway-
> relation
> > >> to fix that? I.e. group several ways for labeling-intentions?
> > >>
> > >> I'm no expert on the inner workings in either of the renderers, but
> to
> > >> me it sounds like a quick fix to a small annoyance. If someone that
> > >> knows the renderers could either agree or disagree, I'd be happy
> anyways
> > >> (well, obviously happier if they agree :)
> > >
> > > See
> > >
> >
> http://wiki.openstreetmap.org/index.php/Relations/Proposed/Collected_Ways
> > >
> > > AFAIK this isn't rendered at present, so for the time being the names
> > > would have to appear on the ways themselves as well if you want to see
> > > them, but in principle, a renderer could take note of this, and if it
> > > becomes a widespread idiom, no doubt they will.
> >
> > I think that is a chicken and egg scenario. I think the renderers (and
> > namefinder) need to support it before people will start using it. Then
> > very quickly we could move all names (and refs and highway types...) to
> > relationships, and we would have a much cleaner data structure.
> >
> > Lots of wierd cases where part of a road has more than one ref, more
> > than one name, or more than one of any other property go away - the
> > relevant ways just become a member of more than one relationship.
> >
> > Personally, I believe that most tagging should be on relationships not
> > ways. Only small physical things like layer, bridge and tunnel should be
> > specified at a way level.
> >
> > Robert (Jamie) Munro
> >
> 
> I think this is one point where the different data clients or consumers
> have
> different preferences. To my mind, you've got it backward. The "small
> physical things" like bridges and tunnels are the parts that should go
> into
> relations, because they have nothing to do with the physical continuity of
> the way. A routing app does not care about bridges and tunnels. However,
> your perspective is probably one of rendering, which would prefer to see
> the
> ways chopped up at bridges and tunnels.
>

I agree we should not be chopping ways for bridges etc, or indeed for speed
limit changes etc.

Will we not in the end be using a combination of 'Collected Ways' (for big
things) and 'Segmented Ways' (for smaller things)?

I suggest that we will settle on a balance between tagging the actual ways
and the relations by experimentation.

I can image using creating a way for a section of road with a particular
name. I might then make it part of a Collected Way to add a 'ref' to it. I
might then use a Segmented Way to assign lower speed limits to certain
sections and a bridge relation to say that it goes over another road.

Time will tell but we are now at that awkward stage where we need both data
and tools that use the data and allow it to be edited elegantly. Possibly
this is exactly the time to thrash out the tagging protocols and to then
encourage implementation.

An important aspect will be to get the renderers etc to handle inheritance
of tags from relationships. I suggest that the precedence is as follows:

A tag on a 'segmented way' has highest priority
A tag on the way itself is next
A tag on a collected way is used where it is not overridden by a tag of the
same name on the way itself or on a 'segmented way'.

One would then be able to assign a 40mph speed to the road as a whole and
then add lower speed limits where it went through each village.

Can I suggest various people commit to develop different areas of data to
support the emerging standards to allow tool-makers to develop and check
their tools out?



Regards,


PeterIto