[OSM-talk] OpenStreetMap routing service

[Ben Laenen]

On Monday 08 September 2008, Nic Roets wrote:
> If the community is capable of agreeing on one worldwide list of
> defaults, then yes. I suggest our defaults be constructed so that we
> need to retag a minimum of roads, e.g. by looking at TIGER
> classifications. Mappers will learn this short list and will be able
> to map parts of foreign countries.
>
> But I think they won't, in which case we need to tag everything
> explicitly. Tagging things explicitly is how you create high quality
> databases.

But then you get roads tagged like this:
foot=yes
bicycle=yes
moped=yes
motorcycle=yes
goods=yes
hgv=yes
psv=yes
taxi=yes
bus=yes
motorcar=yes
etc.

Explicit tags can be a good thing, but I prefer to omit these for now if 
it follows the default rules of that country.

and furthermore, you can get into trouble because country X has two 
classes of mopeds, country Y has no concept of hgv or psv etc. It 
doesn't make sense in 99% of the world to split up between two moped 
classes, so that shouldn't be done at world-wide level.


IMHO, OSM tags should work like this: there's a defined set of base 
tags, for highway types or vehicle types, but they don't need strict 
definitions, only fuzzy ideas about what it means. For example that a 
primary road is "more important" than a secondary road.

Then it should be up to the countries to define those tags strictly 
(when to use what, how to tag them, their maxspeed and access rules and 
things like that). And if the global tags aren't sufficient because a 
country has some special cases, then they can extend the tags (perhaps 
provide some "translation" to global tags that come close to its 
definition). Or a country could discard some globally defined tags if 
they just don't make sense.


My opinion is that since all traffic rules are done at the country or 
state level, it only make sense to define the meaning of all OSM tags 
at that country level. That may need more work in renderers and routers 
and everything interpreting the data, but it just seems more logical to 
me.


> It is far more common for newbies to
> misclassify a road and then we at least still have correct tags. An
> example so fresh that mapnik still renders the mistake :
> http://www.openstreetmap.org/?lat=-25.751366&lon=28.198875&zoom=18&la
>yers=B00FTF

Educate newbies then before they start mapping. It's a problem that 
comes up now and then that you see a new person mapping some things 
that he wouldn't have done that way if he read the country page for 
example or a page like 
http://wiki.openstreetmap.org/index.php/Editing_Standards_and_Conventions 
which should be updated regularly and be mandatory reading material 
before anyone starts mapping IMHO. That would certainly have made me 
spend less time fixing someone's unconnected junctions for example.

> "(b) it fixes the problem where someone might not be familiar enough
> with the traffic rules so he doesn't know for example that
> pedestrians aren't allowed and doesn't add that access tag"
>
> Mappers will still be allowed to omit tags.


> I wouldn't be able to distinguish between and A road and a B road in
> the U.K. and that's what highway=road is for. But I if I see a road
> with cyclists on it, I'll know to tag it with bicycle=yes.

Yeah, but if you just tag what you see you'd be convinced that in some 
places cyclists can use the pavement for example. Seeing a lot of 
people doing it doesn't mean it's allowed.

But I think we can expect mappers in a certain country to at least get 
an idea of the mapping rules in that country, no? It's not that much 
work and you may learn you some traffic rules as well before you start 
mapping that area which can reduce some unpleasant "suprises" on your 
trip...

Greetings
Ben