OpenStreetMap

This diary explores the problems with the OSM Wiki entry for service=driveway, and why is there a need for service=driveway2. In the process, it also describes the headaches encountered while trying find a optimal solution. However, deeply entrenched opinions makes this process untenable.

Need for clear indication of classified service way type

The topic of discussion will need to begin on how OSM tags are designed in an hierarchy, which begins with initial key-value pair of highway=service, followed by classifier tags service=driveway or service=parking_aisle. Note that highway=service without any service=* does not tell the data consumer if this is simply a minor way that is un-classified (not yet assigned a classification type such as parking_aisle or alley or driveway), or if this minor way is an implicit driveway. In general, a data consumer will assume that the way is un-classified, instead of assuming that it is an implicit driveway.

The classifier issue is the first issue present in the service=driveway Wiki entry - that suggests that minor ways linking from major roads to parking_aisles should be tagged as an implicit driveway with no classifier tags.

  • This means that data consumers are unable to determine if the way segment is un-classified, or if it is indeed an implicit driveway when such minor service way connected to parking_aisles are encountered..
  • Similarly, a user whom just want to quickly add minor service roads but does not perform classification will mean that data consumers cannot identify if such ways are classified or un-classified, and therefore prevents accurate assessment if further classification is required.

Definition of driveway

This leads to the second problem of what is meant by a driveway, and what is the problem with the Wiki entry. The problem lies with the “classical” definition of driveway, which in the early days of OSM defines it as “a short private road that leads from a street to a person’s house or garage”, sourced from the “American” section of Cambridge’s dictionary. Oxford/Lexico US likewise provides a similar definition “a short road leading from a public road to a house or garage”. This was the original argument by some Wiki participants on what is meant by driveway - mainly based on US English dictionary definitions.

Nevertheless, let’s dive into history where driveway was first defined - ironically in the US California Streets and Highway Code (SHC), which first introduces the driveway term in 1971. In this law stature, driveway “means a paved portion of a public street providing an unobstructed passage from the roadway to an offstreet area used for driving, servicing, parking, or otherwise accommodating motor vehicles.” This part here is significant because this definition predates the introduction of UK highway codes, and describes general use of minor ways as driveways. Note that California laws are very influential in the auto-industry, and drives formation of rules across the US and beyond (impacts globally). Moving to UK law statures (remember the part of OSM being of UK origins?), the term driveway is itself not defined in the law statures. This strongly suggests that usage of this term is strongly linked to “American” usage of this term, which is something pushed by members of the OSM US chapter.

We will further focus on the UK definitions of the term, which is likewise defined in Cambridge’s UK definition as “a private area in front of a house or other building onto which you can drive and park your car”, and similarly Oxford/Lexico UK’s definition of “a short road leading from a public road to a house or other building”. Note that in both UK dictionary definitions, it never suggests that driveway only links to residential properties which routing engines should avoid routing through, but instead is a more general definition similar to the California SHC where it is a minor way going to any property (I.e. no destination access to residential as contrast to US definition).

The difference in understanding is also why despite the Wiki proclamations of such “rules”, there are a large group of users whom are still tagging wrongly - being that they follow the UK definitions or the SHC definitions instead of the US definitions that should be more appropriately tagged as service=residential_driveway (which should apply a routing penalty). This incorrect Wiki definition is the second problem of the existing service=driveway page, where the optimal and most generic classifier used in SHC cannot be defined under service=driveway tagging.

Implications

Implications to routing

Maybe you may get the implications to routing at this point, but what is meant by service=driveway in the existing Wiki is meant to apply routing penalties such that routing engines do not route through such driveways. However, this is of the mistaken belief that these ways mostly represent destination minor ways whereby most of such ways are of private access to residential buildings. The more correct representation for such ways is the more appropriate service=residential_driveway, to ensure routing penalty is applied to such ways while allowing service=driveway to be routed through at a significantly lower speeds. This is because in most cases, such minor ways are meant to be publicly accessible to gain access to a property, but is not of a major road type such as highway=residential/unclassified which is a major road class.

Implications to “correct tagging” of driveway (using SHC/UK definitions), and proper use of classifier service=driveway

Well, so far so good, but why is there a need to change the current behaviour? Well, other than people tagging using both types of definition, there are often zealous users whom attempt to “correct” such incorrectly tagged ways such that any classifiers such as service=driveway on minor ways linking to service=parking_aisles are removed.

In essence, useful information detailing the minor way type is removed to an un-classified state, leaving the data consumer left guessing if this is an implicit driveway, an un-classified driveway or a residential_driveway as in problem (1). When you attempt to stop them from removing such details, such users typically will just refer to the Wiki as the source-of-truth, despite that fact that the Wiki itself is also not authoritative and is simply meant to document tag usage for the goal of harmonising tags used. This causes problems whereby correct tagging by natural language understanding falls prey to differently defined Wiki entries.

Proposal

Proper Resolution

The most correct way to resolve this problem is to tweak the Wiki to…

  1. Remove the restriction whereby ways connecting minor ways to parking_aisles should be tagged without the service=driveway classifier. In general, it should be assumed that highway=service without service=* means that the way is un-classified but of minor way, and highway=service with service=* are correctly classified ways.
  2. Introduce service=residential_driveway for the purpose of private access and destination driveway to residential properties, which is what Amazon and various entities likes to map. In this way, service=driveway can take on its oldest SHC and UK origins definition which is minor road that is meant for motor vehicles and public access, but does not satisfies other classifications such as service=parking_aisle/alley/e.t.c.

Status Quo

In the absence of such changes, we can only continue to tag the SHC and UK definition of driveway as service=driveway2. However, one will typically still run into zealous mappers whom insist on removing classifier tags such as (1).

Notable Problems when discussing with community

Community

The general sense of trying to have a fruitful discussion with various communities is don’t bother. Despite trying to argue the merits and cons, you will have people insisting on schematic definitions (insisting that US dictionary definitions are correct), insisting that other usages of driveway is wrong, insisting that the definition is local/globally used with consensus reached (and thus dismissing local concerns which actually arise from problems with the Wiki entry), insisting that SHC/UK definitions do not matter, do not recognise the problems faced due to a non-authoritative Wiki entry, and in general dismisses your arguments.

Unprofessional behaviour

Worse, there is one OSM Wiki administrator exhibiting appalling and un-professional behaviour whom insist that this is a user problem. Not only that, this OSM Wiki administrator had to put in his personal opinion by stating that the solution of documenting tag use in the Wiki is the proposal one of user (and that is me). This is something I am proud of, because we are following the process to document tags use instead of trying to fix the service=driveway Wiki page which we have already accepted will be left in the US definition. Indirectly, this becomes a form of personal attack which is uncalled for.

Discussion

Comment from SomeoneElse on 1 September 2021 at 00:53

Oh dear. It does seem that this diary entry comes from a place of pain.

As a native English English speaker, I’m not sure I recognise your “UK” definition of driveway - and anyway, even though OSM tends to use British English rather than the alternatives, there are plenty of exceptions - such as “city” - where OSM generally (and OSM communities seperately) have come up with a definition that doesn’t just come from one dictionary.

If I understand what you’re saying, it’s that “service roads linking parking aisles to roads that are not service roads should be tagged as service=something”, and that “something” can’t be “driveway” because that is used for something else? If so. I’d spend a bit more time explaining why that is a problem - what real-world tasks won’t be possible without this clarification?

I don’t think that’s entirely fair to say:

In the absence of such changes, we can only continue to tag the SHC and UK definition of driveway as service=driveway2.

when a quick look at https://taginfo.openstreetmap.org/tags/service=driveway2#map (and who’s tagging what) suggests that “we” in that sentence is referring to mostly just one mapper - you.

I’m also not sure that I can see the “appalling and un-professional behaviour” that you link to either (in fact I had to guess who you might be referring to based on which of the linked page’s edits was a wiki admin).

Comment from Retired Account on 1 September 2021 at 01:16

Why is there a need to mention “one mapper that mapped 3000 ways” on the Wiki page, where all the page attempts to do is to detail what constitutes the conditions that the tag applies to, and what are the differences with respect to the other tag. The Wiki is always used to clarity tagging conventions, but never before attempting to document what a user did in the main text.

This statement here is clearly a targeted attack on that particular user, and coming from a OSM US board member and OSM Wiki Administrator, this behaviour is clearly upsetting and of poor judgement.

Comment from impiaaa on 1 September 2021 at 01:45

This means that data consumers are unable to determine if the way segment is un-classified, or if it is indeed … when such minor service way connected to parking_aisles are encountered..

This is true. There is a need for a service=* sub-tag to indicate parking lot access roads.

It is also true that the unfortunate naming of service=driveway has caused confusion among those wanting to tag parking lot entrances.

The problem lies with the “classical” definition of driveway, which in the early days of OSM defines it as “a short private road that leads from a street to a person’s house or garage”, sourced from the “American” section of Cambridge’s dictionary.

As I see it, even though they look like it, OSM tags are not English, so an English dictionary is irrelevant. For example:

Nevertheless, let’s dive into history where driveway was first defined - ironically in the US California Streets and Highway Code (SHC), which first introduces the driveway term in 1971.

When writing a proposal, referring to laws as evidence of a feature is a good idea, but not to define the use of a pre-existing tag. OSM tags are not strictly defined based on laws.

insisting that SHC/UK definitions do not matter

That is correct.

This is something I am proud of, because we are following the process to document tags use instead of trying to fix the service=driveway Wiki page which we have already accepted will be left in the US definition.

The process for introducing new tags is to go through a proposal. It is designed to generate community consensus, which can then be documented on a wiki page for the tag. Inventing a tag, then using your own use of it to justify documentation, justifying further use, feels more like a rejection of the community and a petty workaround, than collaborating with the OSM community.

I did not want make this accusation with merely correlation and suspicion. So, I made an Overpass query that returns all uses of service=driveway2 and includes the name of the editor that last edited that way. I then tallied up the numbers of each user, and found that JaLooooNz edited 3029 instances of service=driveway2. The other 14 users that used that tag used it a total of 29 times. The wiki page for service=driveway2 was also first created by wiki user JaLooNz.

this becomes a form of personal attack which is uncalled for.

I have not seen any personal attacks (on either side, luckily), and I don’t understand how documenting tag history constitutes as one. If you feel comfortable, I’m sure many would appreciate a further explanation, or feedback to make communication more accepting. Personally, I would consider “unprofessional behavior” to include things such as ignoring tagging guidelines, and inventing community consensus.

I would be happy to add my “yes” vote to a well-thought-out proposal for a service=* value for what you consider a “driveway.”

Comment from Retired Account on 1 September 2021 at 02:51

The part on adding new entry definition occurs only after the discussion on the service=driveway talk page. Unfortunately, it is not possible to find a better definition for a generic term other than driveway2, as the intention was to trace to the SHC definition so that this broader definition can be used as the classifier tag.

The initial proposal page for parking_aisle and driveway never included statements on issue (1), the changes for (1) was also added directly to the Wiki without any discussion on Wiki talk page/proposal process in 2016/2017. People should not be expecting different process where in the first place, this was never discussed on the talk page when implementing the changes for (1) in 2016 unlike when the service=driveway2 page was introduced following discussion. The more straightforward resolution is to simply remove the part that was never approved, which is the proper resolution steps in the proposal. However, there is great resistance to changing the status quo, and that results in the service=driveway2 tag documentation following a very detailed and comprehensive discussion, unlike the change that occurred in 2016/2017.

Comment from Richard on 1 September 2021 at 10:26

There is so much bizarritude in this diary entry I really don’t know where to start.

service=driveway is predominantly used in OSM to mean a private, cul-de-sac access to a house or other property. That’s what driveway means in British English, and British English is the lingua franca of OSM. Californian highway codes have nothing to do with anything.

If service=driveway is being used to tag the entry to a public car park, that’s a tagging error. The entrance to a public car park isn’t a driveway so we don’t tag it as a driveway.

A routing engine should not need to consider service=driveway as a through-route. In practice, however, mappers make mistakes and sometimes tag through-routes with service=driveway. Therefore it’s sensible for the routing engine to just apply a large penalty to it. This is exactly what I do at cycle.travel, because I found that blocking service=driveway entirely caused occasional failures on the entrances to “trailheads” (as the USians say) and pedestrian/cycle ferries.

The end.

Having a tag called “driveway2” is on the scale of “I can’t even” or “not even wrong”. Please stop.

Comment from yvecai on 2 September 2021 at 17:36

While I used to wholeheartedly agree with the same definition as yours, Richard, lately I’m more inclined to tolerate the other one, widely used, de facto, undocumented : “like a residential, but [rendered] thinner”. Once you come to accept that, further discussion become quite useless.

Comment from Mateusz Konieczny on 12 September 2021 at 12:40

Why is there a need to mention “one mapper that mapped 3000 ways” on the Wiki page

It is normal to do this when tag is used just by a few people. I did it with say https://wiki.openstreetmap.org/wiki/Tag:dual_carriageway%3Dyes when I wrote

Majority of uses are currently by one user in Krakow, Poland, as of late 2020[1]. A typical dual carriageway is not using this tag.

Note that I mapper using this tag and OSM Wiki editor are the same person, me.

In both cases it is not a personal attack.

Comment from Mateusz Konieczny on 12 September 2021 at 12:44

Implications to “correct tagging” of driveway (using SHC/UK definitions), and proper use of classifier service=driveway

Legal definition in California is not sole or primary or only correct driveway definition, OSM tagging is not defined by California law.

The most correct way to resolve this problem is to tweak the Wiki to… (…) Introduce service=residential_driveway for the purpose of private access and destination driveway to residential properties

Redefining heavily used tag would require a very clear consensus, it is NOT an OSM Wiki tweak.

Comment from gecho111 on 12 September 2021 at 15:48

Would definitely like to see driveways to individual residences tagged differently. Some local micro mapping added hundreds of short ways approximately 15 meters in length.

On maps like Strava uses it looks absolutely ridiculous. For the map on my cycling GPS I’ve had to add extra filtering in my mkgmap style sheet to filter out any way tagged as a driveway below a certain length.

Comment from vorpalblade-kaart on 25 August 2022 at 13:47

service=driveway2 recently passed 10k usages, which means JOSM has to look at it.

Considering almost all usage is from a single user (see https://josm.openstreetmap.de/ticket/21396#comment:44 ), and how it is extraordinarily similar to service=driveway from every one else’s perspective, I am leaning towards marking driveway2 as a misspelling of driveway, as most users who do type in driveway2 are likely to have done so by accident.

@JaLooooNz: I kind of get why people don’t like service=driveway on access roads for parking lots, drive throughs, and so on, but you really should have used something that was distinct from driveway. For example, since it seems that the sticking point was that some people wanted to make certain all highway=service ways had service=* tags and other people did not want service=driveway on access ways to parking lots, maybe something like service=access should have been used instead of service=driveway2.

In the future, can you please go through the tagging list and/or wiki proposal process? Or at least don’t use a tag that can be very easily mistaken as a misspelling for another?

Log in to leave a comment