[Tagging] Tagging-rendering relations

[Christoph Hormann]

On Wednesday 09 July 2014, Daniel Koć wrote:
>
> > My opinion is that the best approach would be to establish better
> > means for people to create variants of the style and present them
> > to a broad audicence.  This would have two effects - first it would
> > allow changes
>
> That would be awesome! However in the meantime we can also make baby
> step in this direction: default "beautiful" map (current "stable"
> one) and default "testing" map, where we can have also all the
> default icons to cherry pick the best approach and use them later in
> the beautiful style. What do you think about it?

This would still require significant additional ressources including the 
workload of managing two separate styles.  I don't think testing is the 
problem here, those involved in the standard style design have testing 
environments.  The key is to enable more people to try out new things 
and allow them to communicate and share their results.

> > Fixed zoom threshold are one of the major problems of the current
> > map style, they are selected to look fine for a certain area,
> > usually the favorite city of the one making the style decision. 
> > Choose a different area where the map scale is different or the
> > geographic setting leads to a different distribution of POIs and
> > things fall apart quickly.
>
> So are fixed highways classification: primary road in rural countries
> can look like a track in developed ones... But it's just the cost of
> trying to make things consistent in global scale, not the specific
> rendering issue and we can't fully get rid of it. Maybe we have to
> make some local rules too, but than we loose a bit of uniformity -
> and we can look for the best balance between those things.

No, you are then mixing tagging and rendering which as i said is a 
really bad idea.  Which roads get certain tags should be based on 
universal, objective rules based on properties of the object 'in the 
field', verifiable by any mapper through observation of the road in 
question.  Making sure these roads are shown in a well readable way in 
the map based on the neutral, verifiable information stored in the tags 
and possibly other data is task of the map designer.  This is often 
difficult since for good results map rendering has to take a lot of 
context into account (like if a road is in an urban or rural setting).  
But rigging the tagging rules to spare the map designer this work (what 
you call 'rethink the tagging rules' based on 'rendering issues') is 
counterproductive since it devalues the data stored in the tags.

To get back to the original topic of this thread - you can of course try 
to make a distinction between hills and mountains through tagging and 
for this to be useful data you establish some prominence threshold.  
Then you say mountains should be displayed at z>10 and hills only at 
z>15 (or whatever) - i can assure you if this works well in the 
Netherlands it won't work in Switzerland and if it works in Peru it 
won't work in Greenland (or the other way round - depending on your 
choices).  Then you create a table on the wiki with distinct rules for 
what to tag as mountain/hill for every country of the world which 
might - if followed diligently by the mappers - lead to a halfway 
decent rendering result in small, homogeneous countries.  But as a 
result the tags are essentially useless to actually tell anything 
substantial about the peak in question.

I am sorry if this sounds like a rant but there are simply so many tags 
used a lot but completely useless in terms of informational value 
exactly because of this.  Please just make sure you do not fall into 
this trap with your peak=* concept.

-- 
Christoph Hormann
http://www.imagico.de/