[OSM-talk] Advanced highway tagging

[Gervase Markham]

Jeffrey Martin wrote:
> I'm not exactly sure what the solution is though.

I'm new to OSM, so I may well be reinventing the wheel here, but it 
seems to me that there are three key things you want to know about a 
road, for various common uses of maps:

1) Its official classification (in the UK, M, A, B, C etc.)
2) Some idea of how "fast" it actually is (for routing)
3) Its importance to local travel (which verges on being a rendering
    issue)

Now it seems to me that "defaults" for 2) and 3) can be deduced from 1), 
and therefore we only need to specifically give information about 2) and 
3) when they diverge from what you might expect. For example, in the 
centre of Oxford, a road which is officially an A-road has a "buses 
only" gate.

So I would suggest that whatever tag is ubiquitous (which seems to be 
"highway") is used exclusively for 1), and then there is a possibility 
of adding additional tags to alter the defaults for 2) and 3). This has 
the advantage, noted by Sven, that all mappers make the same decision 
about a particular road.

In the case of 2), it's things like speed limit information and number 
of lanes. We do not want a tag "summarising" this information, because 
the conversion of this data into a single renderable value ("colour") 
should be a renderer algorithm. If it gets the wrong answer, either 
improve the tagging or fix the algorithm.

However, in the case of 3), it's different, because figuring out 
importance is not a case of applying a numerical algorithm to the way or 
node data - you have to take a "wider view". So we have a choice: either 
force all renderers to figure out "importance" from other data ("Is 
there another bigger road between these two 'town' nodes?") or to add a 
"suggested importance", which the renderer should feel free to override 
if it does have more sophisticated algorithms. I favour the latter. 
Let's prefix it with "renderer:" if people don't like using the main tag 
namespace.

I completely agree we shouldn't encode "zoom level", and I also agree 
that renderers should be free to discard this value if they have a 
better algorithm. But I think it's rough to force all renderers to 
implement such smarts, when these decisions are much more easily made by 
a human.

Regardless of what's decided, it seems to me that OSM should be long 
past the stage where the meaning of core tags gets redefined on a 
regular basis. If something as core as road labelling is _still_ in 
flux, let's get whoever needs to agree together and nail it down.

Gerv