[OSM-talk] Segments with 0m Distance

Simon Hewison simon at zymurgy.org
Sun Apr 23 20:03:54 BST 2006

Jim Ley wrote:
> "Tom Carden" <tom at tom-carden.co.uk>
> On 4/16/06, Jim Ley <jim at jibbering.com> wrote:
>>> There's one place near me which is a one way street for 0 distance - 
>>> well
>>> it's not a one way street, there's just a no entry sign preventing 
>>> access
>>> when travelling in one direction, it's 2 way traffic everywhere.  
>>> I've not
>>> got it in OSM yet (I don't go there often enough 'cos it's no entry, 
>>> and at
>>> the top of a very steep and inconvenient hill)  I think that would be 
>>> best
>>> modelled as a no distance segment in a way, but I could be wrong.
>> Interesting - that's why initially I said yes but only if it has no tags.
>> Still - I'm not convinced as to why this shouldn't just be a tag on
>> the node at the no entry end... perhaps a noentry key with a segment
>> id?
> I think the routing people need to decide on that one - I'd just been 
> imagining them preferring not to have to spend too much time checking 
> the tags on every single node?

Routing decisions have to look at the tags on every single segment and 
every single node.

For instance, there could be a single node which represents a low bridge 
at 3.5 metres maximum height, and the user's defined his vehicle is four 
metres tall, and a way might represent a road that is hundreds of 
kilometres long, and there's no way that a single low bridge should 
affect the whole length of the way.

In order to simply find one of a number of possible routes, it barely 
needs to look at any tags, but once it's narrowed down the list of 
possible routes, it makes sense then for it to carefully examine each 
suggested route to check for things like low bridges and the like, 
before presenting it's choices to the end user.

A routing engine should think of ways as merely a method of obtaining 
high level information from the tags about segments that are part of 
that way, and to simplify written instructions, though it does make 
sense to attempt to stick to a signposted route since this makes the 
turns easier to follow.

If the routing algorithm was dumb enough to just try to stick to ways, 
it would end up like Google Map's routing algorithm, which sticks at all 
costs to long distance roads and motorways, even through there are far 
more direct routes that can cut a long distance. (eg on 
maps.google.co.uk, route from "Riseley, Berkshire, RG7" to "Guildford")

Simon Hewison

More information about the talk mailing list