[OSM-talk] Landuse areas etc. abutting highways

Tue Oct 6 17:56:31 BST 2009

Chris

Despite the well-argued views of a minority, I am persuaded by the equally well-argued views of the (considerable) majority who favour option (b).

That is not to say that there isn't room for using a bit of common sense! I wouldn't divide up Delamere Forest into individual areas bounded by paths etc. - the paths in a sense form part of the forest landuse - but I would probably divide a residential area with, say, a major road going through it and would certainly divide landuse=farm either side of a road, for example, if I knew that it was a different farm on either side.

Like everything else in OSM, it all a question of judgement!

I asked the original question from a neutral standpoint but - in the light of the responses have now developed a preference for option (b) - with exceptions.

Of course, nothing is ever final ...

Mike Harris

> -----Original Message-----
> From: Chris Morley [mailto:c.morley at gaseq.co.uk] 
> Sent: 06 October 2009 15:46
> To: talk at openstreetmap.org
> Subject: Re: [OSM-talk] Landuse areas etc. abutting highways
> 
> Mike Harris wrote:
> > Thanks to those who responded to this thread. Advice 
> gratefully received.
> > 
> > There seems to be a clear majority preference for option (b)
> > - the more detailed approach that avoids superimposing 
> boundaries of 
> > areas (and their nodes) on an adjacent way (and its nodes).
> > I fully understand the two caveats:
> > 
> > 1. It is only worth being precise if there is precise data 
> available.
> > 2. There are a few exceptions where, for example, the 
> character of the 
> > adjacent area has access features more like that of a normal linear 
> > way
> > - the pedestrian area is a good example.
> > 
> > I am persuaded that the advantages of forward compatibility and a 
> > higher standard of mapping justify my small efforts (where 
> I have good 
> > GPS data)
>  > in separating out superimposed areas/ways and using option (b).
>  > I am particularly pleased to receive support for splitting 
> single large  > landuse areas (e.g. =residential or =farm) 
> that cross large numbers of ways.
> 
> Let me encourage you to use option a), based on the reasoning 
> of Frederik Ramm.
> 
> In detailed mapping, everything is an area way which share 
> nodes with its adjacent areas. When roads etc. are linear 
> features, it means they have *indeterminate* width and the 
> only non-arbitrary representation of this in an editor is for 
> the width to be zero, with adjacent areas on both sides 
> sharing the nodes - option a). This makes it consistent with 
> the detailed modelling approach. I would look at the linear 
> road etc. as being, not a centre-line, but an indeterminately 
> wide structure comprising the road surface, sidewalks, verges 
> etc. up to a boundary (which in the British countryside would 
> often be a hedge.) By mapping with option a) you are saying 
> that the golf course, say, comes up to the road's boundary 
> hedge but that you haven't specified exactly where that is. 
> If you do know, you are into a detailed mapping approach. If 
> a linear road is still used then it would now be interpreted 
> as a centre-line, as is sometimes done with rivers.
> 
> Since I map in the same are as you, I suspect that in most 
> cases you do not have enough information to use the detailed 
> mapping approach. 
> Even with arial photography we have available, poor 
> resolution and interference from tree cover and shadows often 
> does not allow the separation between the hedges to be very reliable.
> 
> Editor support for ways sharing nodes is certainly poor, but 
> as with inadequate renderers, we should improve them rather 
> than adding artificial data (arbitrarily positioned 
> structures) into the database.
> 
> Landuse areas which cross a large number of ways are very common. 
> Surely you don't intend to divide say, Delamere Forest, into 
> a large number of separated areas separated by the paths and 
> tracks? When you do need to do it, separating an area into 
> two at a road is certainly laborious and maybe somebody 
> should build a JOSM plugin to do it.
> 
> Chris
> 
> > 
> >> -----Original Message-----
> >> From: Martin Koppenhoefer [mailto:dieterdreist at gmail.com]
> >> Sent: 05 October 2009 15:52
> >> To: Marc Schütz
> >> Cc: talk at openstreetmap.org
> >> Subject: Re: [OSM-talk] Landuse areas etc. abutting highways
> >>
> >> 2009/10/5 "Marc Schütz" <schuetzm at gmx.net>:
> >>>> 2009/10/5 "Marc Schütz" <schuetzm at gmx.net>:
> >>>>>> But a) could be used as acceptable temporary solution until 
> >>>>>> someone with better information (like having aerial
> >> photography)
> >>>>>> remaps it as
> >>>>>> b)
> >>>>> Yes, this is basically what I wanted to say. Leave it to the 
> >>>>> mappers
> >>>> whether they want to use a way or an area for a road.
> >>>>
> >>>> it will be much harder to add this detail, if all areas
> >> are merged though.
> >>> Not really. JOSM supports disconnecting ways since a long
> >> time now. But anyway: doing things wrong just to make 
> editing easier 
> >> is not a good thing.
> >>
> >> +1. That's why adjacent landuses (see topic) shouldn't be 
> extented to
> >> the center of the road.
> >>
> >>>>> But with option (b) and a linear way you would have a
> >> gap next to
> >>>>> the
> >>>> road. In the case of landuse, this is not a problem in
> >> practice, but
> >>>> if there is a place, there you need to insert artificial 
> ways that 
> >>>> are not there in reality, just to get the connectivity
> >> between the two objects:
> >>>>> http://osm.org/go/0JUKytHID--
> >>>> which objects are you referring to? parkings usually have
> >> those ways
> >>>> (for crossing the sidewalk) so they won't be artificial, and 
> >>>> pedestrian areas are the exception I mentioned above.
> >>> Look at the google sat image:
> >>>
> >> 
> http://maps.google.com/maps?f=q&source=s_q&hl=de&geocode=&q=bayreuth&
> >> s 
> >> 
> ll=37.0625,-95.677068&sspn=59.856937,107.138672&ie=UTF8&hq=&hnear=Bay
> >> r
> >> 
> euth,+Bayern,+Deutschland&ll=49.946316,11.577148&spn=0.000754,0.00163
> >> 5
> >>> &t=k&z=20
> >> That's the mentioned pedestrian area. I agree with you here.
> >>
> >>> Mapping it the way it is done there does not really make
> >> sense: Either the exact geometry is important for you, then you 
> >> should convert both the plaza and the road to areas. Or it 
> isn't, but 
> >> then there shouldn't be a problem with extending the plaza 
> so that it 
> >> borders to the road.
> >>
> >> +1. but that's still pedestrian areas / highway areas. In 
> these cases
> >> the areas _do_ connect to the road.
> >>
> >> cheers,
> >> Martin
> >>
> >>
> >>
> > 
> > 
> > _______________________________________________
> > talk mailing list
> > talk at openstreetmap.org
> > http://lists.openstreetmap.org/listinfo/talk
> > 
> > 
> > 
> ----------------------------------------------------------------------
> > --
> > 
> > 
> > No virus found in this incoming message.
> > Checked by AVG - www.avg.com
> > Version: 8.5.420 / Virus Database: 270.14.4/2416 - Release Date: 
> > 10/05/09 18:23:00
> > 
> 
> 
> 
> 