[Tagging] [Talk-us] Freeway exit tagging

Paul Johnson baloo at ursamundi.org
Fri Aug 26 17:48:32 UTC 2016

On Fri, Aug 26, 2016 at 11:31 AM, Georg Feddern <osm at bavarianmallet.de>

> Am 26.08.2016 um 14:18 schrieb Kieron Thwaites:
>> I can, however, see the rationale behind tagging "none;slight_right",
>> as well as tagging nothing at all, and as such, I think that this is
>> an issue that we need to find consensus on.  That said, I believe Paul
>> is quite correct with his statement that machines "need to be told
>> about these restrictions in order for them to be able to provide
>> useful feedback from it" -- something that isn't explicitly present
>> (or maybe not even implicitly so) but appears obvious to a human on
>> the ground isn't necessary obvious to a machine.
> Please do not take it personally - I just toke your answer to hook on,
> because you agreed on "machines need to be told about these restrictions".

No problem, I'm willing to take it at face value for clarifying my position.

> May be I am a bit on the devils advocates side here ... ;)
> 1. Where are there any restrictions? There is no solid line between the 3
> lanes. ;)

Right, but we're not talking about change:lanes=* here.

> 2. If you want to give the machine any advice, you should take
> "through|through|through;slight_right"
> because
> 3. "none|none|none;slight_right" does not give any advice for the both
> left lines - they still could be considered to take the exit.

In the physical sense and a "no cop no foul" sense, sure, under ideal
conditions.  In practice, depending on where you're at, it's going to
garner some variation or combination of improprer lane change, or not
turning from the nearest lane, or cause a collision.  We shouldn't be
helping Mayhem play GPS in https://www.youtube.com/watch?v=4h0Qvc6_MfQ if
we can avoid it, even if it still is the driver's responsibility to not be
an idiot (or trust a driverless car to the data without supervision).  It's
still not helpful or going to gain many new end users or contributing
mappers to make it easy for data consumers to give them incorrect
information due to incomplete information just because the DOT left out a
sign or a pavement arrow.

> Well in reality that is the legal situation here - you just have to take
> care for the traffic on the lines.

At the risk of being moderately offensive with the stereotypes, this does
remind me of a Family Guy bit about driving...

But in this common case (standard single lane exit) I still do not see any
> necessarity for any advice to the maschine (or the driver), that if the
> route takes the right road one should use the rightmost lane ...
> Same situation with solid lines definitely need case 2 - because even you
> should 'implicitly know' to take the rightmost lane there is another point
> where you already _have_ _to_ be on the rightmost lane - the maschine needs
> this advice to announce it appropriate.
> But may be I am a bit too old and have driven too long with my own eyes
> and head - and without a navigation assistant. ;)

This isn't about humans driving in familiar territory, it's about
facilitating automated driving or providing cues for drivers unfamiliar
with the area how to best navigate a new region.  Oddly enough, on
particularly well lane-mapped areas, I'm actually quite surprised how close
Osmand is to lane choices a local with extensive experience with an area
would pick.  Omitting things or leaving "none" in there generally throws
off some oddities (like suggesting a spot where there's two marked left
turn lanes, three lanes with no markings and a right turn lane with an
arrow that you can turn left from the five leftmost lanes or right from the
four rightmost).  Rather than expecting data consumers have to have
programmatic assumptions for every jurisdiction, imparting this information
even if it's not explicitly signed or posted yet only debatable in the most
farcical of situations, should be considered ground truth for the purposes
of mapping and not just killing potential developers interested in
providing lane guidance.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20160826/94525536/attachment-0001.html>

More information about the Tagging mailing list