[Talk-ca] CanVec natural=land tags and untagged ways

[Adam Dunn]

The natural=island tag that Daniel is referring to used to be applied
to the way of the island. This is the old way of doing things
(pre-Canvec 7). I think the natural=land tag that Samuel is referring
to is a single node at the centre of the island (Canvec 7).

The natural=land node is there for the purpose of retaining toponymy
(naming). Many islands don't have names and you can just delete the
node, but some of these nodes will have the name of the island, so you
should either keep the node or transfer the name of the island over to
the island's outer way.

For water body relations (not coastal), it is sufficient to have just
a closed inner way polygon; you don't need a natural=land tag (or any
other tags). I'm not that experienced with coastal tagging, but I
think having a way going the correct direction around the island and
tagged as natural=coastline is how to tag an island in the ocean/sea.
One shouldn't need a natural=land in that case either (in fact,
according to the wiki, having natural=land as the sole tag on a costal
island is not the correct way of doing things [1]). The two cases
where natural=land is required is when the island is only a node (too
small to be a way polygon), or when you aren't using relations and
need to have an island way polygon (but this is obsoleted by using
relations).

I thought the tagless "ghost" ways were a byproduct of how JOSM
deletes relations, I didn't know it was part of the Canvec export's
construction. They can be tossed.

Adam

On Wed, May 18, 2011 at 7:25 AM, Bégin, Daniel
<Daniel.Begin at rncan-nrcan.gc.ca> wrote:
> Hi Samuel,
>
> about a year ago, I removed natural=island ways from the Canvec data. Unless
> I'm confused (it appends sometime !-) it was applied for Release 7...
>
> The problem was that islands were/are overlaying all other features on
> rendering, including corresponding natural=wood features (ie : wooded
> islands renders white spot instead of green)
>
> If you still have natural=island features you should be in an area where the
> Release 7 could not be produced (about 30 files for the country)
>
> About the ghost ways, it was decided to create the Canvec product that way
> to ease partial/layer import (for example, import hydrography without wooded
> areas). However, once you have modified the data to merge both features,
> I don't see the need to keep ghost ways.
>
> Regards,
> Daniel
> ________________________________
> From: Samuel Longiaru [mailto:longiaru at shaw.ca]
> Sent: May 18, 2011 09:42
> To: talk-ca
> Subject: [Talk-ca] CanVec natural=land tags and untagged ways
>
> Good morning everyone,
>
> I've been working for the last couple of months importing Canvec data into
> south-central BC and have almost completed the eastern half of 92I.  I also
> have been lurking on the MkGMap list and one of the comments there today got
> me thinking that maybe I've been doing something wrong.  Just wanted to get
> some comment here if I might.  I can go back and fix things if need be.
>
> The procedure I have been using for importing is essentially a reflection of
> what I would normally do should I be mapping an area from scratch.  I select
> a feature like wood, wetland, water, etc. from my CanVec data layer, check
> it against the existing OSM, merge where appropriate and delete the feature
> from my CanVec data layer so I can keep track of what I have done.  At the
> end of this process, I am usually left with a couple of things in the CanVec
> layer which I discard.  For example, after merging "wood", I delete it from
> the CanVec layer and in many cases am left with another untagged way that
> follows the wood boundary.  This way has no tags at all and is not part of
> any relationship.  As it normally would not be present should I have just
> traced the wood using Potlatch or JOSM, I delete it and do not import it
> into OSM.  I have also been ignoring the natural=land tags that appear on
> islands in lakes.  I have not been importing this tag since if I understand
> things correctly, it is sufficient to have islands tagged only as inner
> members of  relationships.   As a check, I have gone back and examined the
> rendered OSM maps, and all wood and islands are rendering correctly.  I have
> also imported some of my imported CanVec data into my Garmin Nuvi through
> Lambertus's site and all render correctly as well.
>
> In response to a query on the MkGMap list as to why oceans were not
> rendering as blue on someone's Garmin (I have this problem too by the way)
> the comment was made that islands needed to be tagged as natural=land.  I'm
> not sure that is actually the case but it did get me thinking about the
> island tags I have been discarding and the other superfluous CanVec data I
> have also been tossing.
>
> Is it OK to toss these natural=land tags?  And what is going on with these
> ghost ways that appear under under the boundaries to wooded areas?  OK to
> toss them as well?
>
>
>
> _______________________________________________
> Talk-ca mailing list
> Talk-ca at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-ca
>
>