[OSM-talk] oneway=1

[Arieh Skliarouk]

See inlines below:

On 10/2/07, Chris Fleming <me at chrisfleming.org> wrote:
> > I am sure commercial mapping companies have some kind of agreement
> > with authorities regarding directions change. Until OSM has something
> > similar, we can't rely on this information to remain static.
> >
> > Direction tags on roads wider than certain width, can be relied upon,
> > though. For narrow streets the direction needs to do deducted
> > statistically.
> >
> > I see following algorithm as one that can work:
> >
> > Initially, mark all narrow streets as having "questionable" direction,
> > with allowed direction of the one that submitter have travelled on it.
> >
> If both directions are permitted and used then this is a two way street,
> if only one direction is permitted (and used) then it's a two way
> street. I'm not sure how this can be questionable?

When the server have not seen GPS-traces from two-ways street in
known-to-be-allowed direction, but saw several traces in the other
direction, that can raise "questionable" flag for the "disappeared"
direction, that might trigger seeking acknowledgment from users that
the street is still two-way one.

> > The client software (one that runs on PDA) needs to be able to notify
> > server that the car just passed the street in previously
> > known-forbidden direction. Once several similar-looking GPS-traces
> > from different users are received by server, the street direction can
> > be recalculated.
> >
> > Similar way, server should deduct streets that have not seen certain
> > direction GPS-traces and mark the street as "questionable" direction.
> > PDA-client should notify the user about questionable status of the
> > street direction and allow clarification of the data by user, with
> > subsequent upload to the server.
> >
> Surely it's better that someone does this manually, rather than trying
> to deduce this from traces?

In absence of manual data, deduction is the best resort. Manual data
will override deduction, but will become void once deduction strongly
disagrees.

> > Other factors that can change in lifetime of a street (off the top of my head):
> > * Whether only public transport is allowed
> > * Trucks are allowed only during certain hours
> > * Cars with dangerous cargo (explosives, fuel, etc) are not allowed
> > * road is under (re)construction, find another route
> > * Only certain maximum weight or height vehicles are allowed through
> > * Pedestrian only
> > * etc
> >
> > What do you think?
> >
>  Yes. But these things don't tend to change that often? A big community
> of local mappers (or even just users who notify mappers?) would catch

You greatly overestimate number of active local mappers. There are
only so much users that will proactively review maps, and much more
ones that would not mind to answer couple questions on their pda (if
asked immediately in the vicinity of the questionable entity).

> these very quickly, probably much quicker that the mapping companies,
> who often get these things wrong anyway. For example the recently opened
> M9 Spur on google is almost right:
> http://maps.google.co.uk/maps?f=q&hl=en&geocode=&q=EH30&ie=UTF8&ll=55.973318,-3.399582&spn=0.040104,0.088491&z=14&om=1
>
> But the new road is Motorway (M9 Spur) not  the A8000. The A8000 is now
> the B800! On OSM it's correct (we also have all the roads crossing at
> the right layers):
> http://www.informationfreeway.org/?lat=55.975530134242675&lon=-3.390307315041101&zoom=14&layers=0000F0B0
>
> I know of dozens more examples where commercial maps have not updated to
> take changes into account, but OSM has...

OSM is created by users for users, not to be on-par with commerical
maps. If we can minimize incorrectness even more, why not?

-- 
Arieh