On 2020-05-25 00:16, Florian Lohoff wrote:
> On Sun, May 24, 2020 at 11:54:02PM +0200, Colin Smale wrote: On 2020-05-24
> 23:16, Mateusz Konieczny via talk wrote:
>
> Can you give an example of such untaggable restriction?
> In the UK there are many small roads signed as "Unsuitable for HGVs."
> Legally you are allowed to drive your 44T truck down there, but you will
> almost certainly get stuck. How do we tell the router?
width? maxwidth?
It needs to be a physical attribute, not a purely legal one. It could be
a combination of road width and bends, or undulations giving a risk of
grounding. For the former, the router would need accurate geometry info
(centre line and width) which is often not present or not reliable in
OSM. For the latter, do we have anything for "risk of grounding due to
dips and humps"?
> It a different attribute than legalese which makes it unsuitable - so
> tag it appropriate.
>
> There are also many roads signed as "No HGVs except for access." It is
> tempting to tag them as "hgv=destination" but that doesn't cover the
> case where you are allowed to follow that route for many miles and make
> several turnoffs IF you "need access". The current definition of
> "access=destination" doesn't allow routers to distinguish between truly
> "first/last segment only" and "its fine if you are going to/from this
> general area".
> Discussion here shows the
> http://www.trucknetuk.com/phpBB/viewtopic.php?f=5&t=140446
> Thats a technical difficulty in the OSM Data model which may fill pages.
>
> At least in Germany a restriction sign is not a "linear restriction"
> e.g. is restricting the whole way. Instead you may not traverse the
> point of the sign. We are currently unable to put this into OSM.
>
> A workaround is to put 2 short oneways on top of each other - one of
> them carrying the restriction - which is in itself a pretty ugly
> solution - and - this does not work for destination.
>
> There are other problems. A destination technically is currently solved
> by increasing the "cost" in the routing graph. So for example for
> every meter on a destination road you may travel 20m on others. Which
> most of the time works pretty well in avoiding the destination roads.
> It has pretty bad side effects which causes the router to try to send
> you out of the destination area with the shortest way even producing
> very long diverts around.
You talk as if all routers are the same.... I accept that such
heuristics are inevitable to choose between multiple possibilities, but
proposing a route that would actually be considered illegal should not
ever happen (subject to data currency considerations).
But still, access=destination does not permit the router to apply
different penalties to the two cases I mentioned.
> Legally this is broken. Legally you may not enter the zone when
> your destination is not within that zone and there nothing like a
> distance based penalty within that area.
>
> So yes - there is a problem - But not within tagging. Its something
> routers need to solve.
Routers cannot work alone, they have to work together with the tagging.
It's not fair to claim it's all their problem. If the tagging does not
represent the nuances required, the router should not be expected to
just guess (at least where the difference is between legal and illegal).
_______________________________________________
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk
