Proposal:Motorway link no default oneway

motorway_link no default oneway
Proposal status:	Rejected (inactive)
Proposed by:	Jojo4u
Tagging:	highway=motorway_link Data item
Applies to:
Definition:	highway=motorway_link without oneway=* does neither mean oneway=yes nor oneway=no. Explicit tagging is strongly recommended.
Statistics:
Draft started:	2015-09-10
RFC start:	2015-10-04
Vote start:	2015-10-29
Vote end:	2015-11-12

Proposal

Strongly recommend explicit tagging of oneway=* on highway=motorway_link.

Define that highway=motorway_link without tagged oneway=* has no implied oneway=yes and also the standard default of oneway=no does not apply. The oneway=* status of such a way would be undefined.

• For rendering purposes ways with undefined oneway should be displayed like the default, i.e. without oneway arrows.
• For routing purposes no recommendation for ways with undefined oneway is made. A provider should decide on it's own considering the documentation history and current data.
• In map editors undefined oneway should be displayed as tagging error. Corresponding tickets will be opened for JOSM/iD/Potlatch.

Rationale

The goal of this proposal is removing the implied oneway=yes on highway=motorway_link from documentation. The following implied default oneway=no is also undesireable and could lead to dangerous situations in navigations.

Whether oneway=yes is implied on motorway_link has a long history, making the documentation about it an outright mess and assumptions on it unreliable.

Some history:

• In 2008 implied oneway=* was added on the wiki page of motorway_link, reverted, and readded.
• In 2007-2008 a discussion about implied oneway=* was divided.
• In 2008-2010 a majority on the talk page was against implying oneway=*.
• In 2013 a section was added that oneway=* should always be tagged.
• In 2015-04 the implied oneway=* was removed from the infobox.
• In 2015-09 I changed OSM Tags for Routing and Highway Link accordingly.

Around the world:

• A check in North America of motorway_link missing oneway=* revealed that perhaps two third of them assume oneway=yes ([1], Exit 15 Paliasades Interstate Parkway Exit 15 Paliasades Interstate Parkway, CA 99 link CA 99 link)
• A check in Europe revealed very few missing oneway=no. Most occurances are roundabouts which have oneway implied (example).
• China and Japan have massive amounts of missing oneway=* with generally assumed oneway=yes (example junction).
• Two-way motorway links are very common parts of Europe (Overpass Turbo Example), as a local mapper without reading documentation I would assume standard oneway=yes.
• "In the US, all exit ramps built to modern (1960's+) standards are one-way[2]"
• "in the TIGER-imported interchanges, there's no correlation between a ramp's actual direction and the direction of the way representing it."[3]

Programs and General:

• OSRM implies oneway=*, while Mapquest and Graphhopper do not.
• OSM-carto does not imply, while JOSM (after some shuffling) still does (since 2008), also ID
• 98% of all motorway_link already have oneway=*.
• We map the world and explicit tagging is preferable over implicit.

Taginfo Statistics

• 97.99% of all motorway_link have oneway=*
• 94.09% of all have oneway=yes
• 3.48% have oneway=no.

Considerung the facts it is clear that our tools and assumptions are already different and thus oneway=* should be explicitely tagged.

Tools to help enforcing the explicit tagging:

• Documentation on wiki
• Tagging as error in editors
• Maproulette challenge(s)
• Community campaigns for explicit tagging (for example Germany)

Examples

• Overpass Turbo Query of empty oneway.

Features/Pages affected

• Tag:highway=motorway_link
• Highway_link#Special_considerations_for_motorways
• OSM_tags_for_routing#Oneway

Comments

First message on Tagging mailing list.

Reqest for Vote.

Please comment on the discussion page.

Voting

• I approve this proposal. --Seichter (talk) 20:45, 29 October 2015 (UTC)
• I approve this proposal. No reason to add an exception to the general rule. The oneway assumption on roundabouts might need attention too. Math1985 (talk) 21:48, 29 October 2015 (UTC)
• I approve this proposal. --Klumbumbus (talk) 22:52, 29 October 2015 (UTC)
• I oppose this proposal. It is very easy to override this default and including the implicit tagging means that the majority of ways (more than 94%—see above) that do in fact warrant oneway=yes are even easier to make. Therefore, the result is an unnecessary burden on the mapper. –(DCTrans)

Explicit tagging is easy to do thus there is no need for any assumptions about links to be made. Here in Thailand there are hundreds of motorway links that allow U-turns and sometimes even allow lower class roads to cross. I explicitly tag this type of link with oneway=no but in cases where mappers forget to do that I would not want routing software to assume anything --AlaskaDave (talk) 03:19, 30 October 2015 (UTC)

That's exactly what this proposal is about: Currently the assumption is that highway=motorway_link is always oneway=yes, unless oneway=no is tagged. Accepting the proposal would mean it's the same as for all other highway=*, except that it's always recommended to explictly tag oneway=no (which is currently neccessary, as it's defaulting to yes). The "recommended" part is only because it's always assumed that it's common that highway=motorway_link should have oneway=yes. --rayquaza (talk) 07:50, 30 October 2015 (UTC)

• I approve this proposal. Okay, got it. I misunderstood the proposal earlier. --AlaskaDave (talk) 23:08, 30 October 2015 (UTC)
• I approve this proposal. oneway=yes is neccessary wherever it's true, and it should be the same on highway=motorway_link. As most highway=motorway_link will have oneway=yes, oneway=no is neccessary to clearly state (i.e. to other mappers) that this is not just a missing tag. --rayquaza (talk) 07:50, 30 October 2015 (UTC)
• I approve this proposal. --Jojo4u (talk) 07:56, 30 October 2015 (UTC)
• I approve this proposal. --streckenkundler (talk) 08:57, 30 October 2015 (UTC)
• I approve this proposal. --Tordanik 09:08, 30 October 2015 (UTC)
• I oppose this proposal. saying that oneway status can be undefined and allowing GPS (routing software) to send cars there is risking to cause accidents --Papou (talk) 11:09, 30 October 2015 (UTC)

The intention of the proposal is to provide clarity and avoid uncertainty. Your assumption about accidents is a misinterpretation. Further, GPS is a positioning system operated by US military and not a routing software.

The first draft included disallowing routing on links with missing onway. The reactions where negative so I removed it. Data customers still decide on their own, so the proposal tries to improve data.--Jojo4u (talk) 18:54, 9 November 2015 (UTC)

• I approve this proposal. --Polarbear w (talk) 11:31, 30 October 2015 (UTC)
• I approve this proposal. --TOGA (talk) 17:26, 30 October 2015 (UTC)
• I approve this proposal. --Paul Johnson (talk) 10:51, 31 October 2015 (UTC)
• I oppose this proposal. Default values are needed by applications. Reasonable defaults also simplify tagging a lot. Undefining defaults is clearly a step in the wrong direction, and you just cannot do that for a feature with 511082 occurences in the database. I disagree with "explicit tagging is preferable over implicit". Do you explicitly tag all roads with access=yes, brigde=no, tunnel=no, layer=0 etc.? --Fkv (talk) 19:23, 31 October 2015 (UTC)

The rationale of the proposal tries to show that there is no "reasonable default".--Jojo4u (talk) 11:50, 1 November 2015 (UTC)

That may have been your intention, but the facts you give in the rationale strongly suggest oneway=yes as default: 94% explicit oneway=yes, most missing oneway tags (as listed in the "around the world" section) also meaning oneway=yes. A reasonable default is a value that is correct in most cases. --Fkv (talk) 14:09, 2 November 2015 (UTC)

• I oppose this proposal. I think that it´s better to have a default for applications which need it --Waldviertlerkartograph (talk) 20:31, 31 October 2015 (UTC)
• I approve this proposal. I have checked some of the 2% missing oneway tags and found about 20% which have assumed as default oneway=no, so this proposal is the best solution for this often misunderstood default value. --WalterSchloegl (talk) 10:46, 1 November 2015 (UTC)
• I oppose this proposal. default makes sense to me, it's easy to override --/al 12:27, 1 November 2015 (UTC)
• I have comments but abstain from voting on this proposal. 3.48% motorway_link have oneway=no, and 3.94% (a bit more) residentials have oneway=yes. So should we also tag all other residentials with oneway=*? But on the ohter hand with this proposal we can easily find the some untagged oneway=no.
  OT: But it is much more important to add oneway to cycleways to allow a valid bicycle routing! (Taginfo oneway=yes: 8.23%, oneway=no: 3.72%) --HalverHahn (talk) 13:53, 1 November 2015 (UTC)
• I oppose this proposal. Plenty of inconsistencies and unacceptable behaviour in the proposal, see these two posts on the tagging list. --Kelerei (talk) 10:12, 4 November 2015 (UTC)
• I approve this proposal. --Michi (talk) 20:25, 7 November 2015 (UTC)
• I approve this proposal. -- Peda (talk) 16:57, 8 November 2015 (UTC)
• I approve this proposal. --User 5359 (talk) 18:16, 11 November 2015 (UTC)
• I approve this proposal. --chris66 (talk) 18:18, 11 November 2015 (UTC)
• I approve this proposal. --Rogehm (talk) 18:37, 11 November 2015 (UTC)
• I oppose this proposal. highway=motorway_link without default for oneway means that router programmers have to abort routing when running into such a way. I don't believe that this happens and therefore the decision of the default behavior passed to the routing programmer. That is really bad! --Hfst (talk) 18:45, 11 November 2015 (UTC)
• I approve this proposal. --Q un go (talk) 19:40, 11 November 2015 (UTC)
• I oppose this proposal. If 98% of motorway_link already have oneway=* then what exactly is the problem? I disapprove of this because if someone really wants to do something about it, the right thing is to write a QA check and submit patches to the editors, instead of starting a vote and then opening tickets that create work for others.
• I approve this proposal. --Christianb (talk) 20:41, 11 November 2015 (UTC)
• I oppose this proposal. A missing oneway is dangerous. Different tagging schemes for motorway_link and motorway are bogus. --Netzwolf (talk) 22:41, 11 November 2015 (UTC)
• I approve this proposal. Not all motorway links are separated streets in reality. They use common ways with both directions! Editors have to sign them with oneway=no for routing availability. This proposal is a better solution, although the dangers of missing oneways exists. --Robybully (talk) 07:08, 12 November 2015 (UTC)
• I approve this proposal. --Garmin-User (talk) 07:59, 12 November 2015 (UTC)
• I approve this proposal. --Nop (talk) 08:27, 12 November 2015 (UTC)
• I oppose this proposal. I think that it´s better to have a default for applications which need it, too. --Klaus-Geo (talk) 11:07, 12 November 2015 (UTC)

If this proposal is accepted, then applications will still be able to assume a default. However, they will know that this is entirely an assumption on their part, and not something a mapper has accounted for. I don't see how that harms them in any way. --Tordanik 10:31, 12 November 2015 (UTC)

• I approve this proposal. --hwb (talk) 10:51, 12 November 2015 (UTC)
• I oppose this proposal. I don't like the concept of turning the default into `undefined` and leaving the decision to the router. In the transition phase this might lead to a bit of chaos. However, I fully agree that the oneway=* should be made mandatory and that editors/QA tools should complain about missing oneway=*. Mmd (talk) 12:06, 12 November 2015 (UTC)
• I approve this proposal. --OPerivar (talk) 17:28, 12 November 2015 (UTC)
• I approve this proposal. --Slhh (talk) 22:58, 12 November 2015 (UTC)
• I approve this proposal. --Protoxenus (talk) 20:42, 13 November 2015 (UTC)

Voting closed

Voting on this proposal has been closed.

It was rejected with 25 votes for, 11 votes against and 1 abstention.

The proposal did not reach the required quorum of 74%.

Analysis

Analyzing the 11 no votes and 2 feedbacks on the mailing list (imho):

• 7 say that a default should be implied
  • 2 explicitely state that oneway=yes should be implied when missing
  • 1 explicitely state oneway=no should be implied when missing
• 2 say that oneway=yes should always explicitely tagged
• 4 say that routing should be given rules
• 1 say that links without oneway should be reduced by QA without proposal shenanigans
• 1 say that undefined oneway is no good idea but requiring oneway is

--Jojo4u (talk) 18:55, 13 November 2015 (UTC)