[OSM-talk] The Highway tags and other junk strikes back

[matthew-osm at newtoncomputing.co.uk]

On Tue, Dec 19, 2006 at 12:00:01AM +0000, Ben Robbins wrote:
> >To sum up your position: Because I support drawing streets wider than
> >they really are, it makes no sense for me to support making gates
> >smaller than they are. Is this a correct interpretation of your
> >argument?
> 
> The point of a street being wider is so its clearer.  So to go smaller is 
> less clear.  Even if the node is to render a sybol or something where the 
> gate falls, that is less clear.  Therefore it seems illogical to state that 
> something should be a node, becuase another segment is rendered wider.  I'm 
> awair that your point is consistent in the extent that your saying that the 
> scaling in realitiy doesnt have to be 'exaclty' the same as on the map.  I 
> would agree to that, but the scaling of gates, seems to be the oposite of 
> your example with the road.

No, for this reason:

You could tag, reasonably sensibly, like any of the following:

a)

*----road----*----road----*
             g
             a
             t
             e
*----road----*----road----*


b)
             *
             g
             a
*----road----*----road----*
             t
             e
             *


c)

*----road----*(gate)----road----*



(a) is like OS Mastermap - it is far too detailed for OSM currently, and we just
can't currently compete in this level of accuracy, in my opinion. Maybe in the
future, when common GPS units have the accuracy and detailed satellite imagery
is available everywhere, but not now.

(b) seems to be the way you are tagging.

(c) is the "standard" OSM convention at the moment.


The argument that you gave (there is an inconsistency about roads being wider
[for readability] but gates being smaller) does not hold when a renderer that
understands gates becomes involved. Current renderers do _not_ understand gates,
and just put an icon over the road. This, I imagine, will be fixed over time.

This is what happens when a road is rendered:


  *--------*

becomes

  ------------

  ------------

it could be scaled to make the map clearer:

  ------------



  ------------


Now, imagine that a gate is a single node (here represented by g):

  *------g------*

The _renderer_ can automatically scale this gate to match the road:


  ------*------
        g
  ------*------

or, bigger for readability:

  
  ------*------
        g
        g
        g
        g
  ------*------


However, when you tag like you do, the renderer has no choice. For a "normal"
scale drawing (whatever "normal" is), you get:

  
  ------*------
        g
  ------*------

but when roads are drawn larger to make them easier to see, you will get the
following, because you have fixed the endpoints of the gate:

  ------ ------
        *
        g
        *

  ------ ------

Similarly, for roads drawn smaller:

  ------*------
  ------g------
        *

So by not following the current convention of using a single node for a gate,
you are causing problems later on for renderers that need to display roads (or
otherwise) at different scales. Besides, as I said earlier, I don't believe that
OSM is currently at the level that we can even think about recreating a
Mastermap type map - let's get the country mapped in a streetmap type way first,
and then add more details later if we can get it accurate enough![1]

But... as others have tried to say; there is nothing stopping you from doing it
your way if you want! Just don't expect everyone else to do the same thing just
because you think you have the best answer!

Happy mapping :)

-- 
Matthew


[1] ...and if we did that, we would have to make sure we were incredibly
accurate, otherwise there would be no point whatsoever. This type of accuracy,
for all intents and purposes, is not currently attainable with off-the-shelf
consumer GPS units. Really.