On one hand, highway=street can be seen as a nice reference to OpenStreetMap :-)
Unfortunately, it also seems confusing to have highway=road that can represent a street and highway=street that can represent a road.
Using street to describe a rural road is certainly confusing and unintuitive. Street implies an urban environment.
- I didn't realize this as a non-native speaker! Thanks for pointing that out. We wouldn't want to replace one confusing value with another! Martijn van Exel (talk) 19:58, 14 July 2021 (UTC)
- Hi, I am the author of the proposal. I have received some similar feedback about the name through different channels, and this is something we need to think more about. Besides the name, do you see merit in the proposal otherwise?
- I don't feel strongly about this: highway=unclassified and highway=residential are not great terms either, but would one option be to just explain more clearly what they mean in practise? --TuukkaH (talk) 16:28, 11 July 2021 (UTC)
- Looking at the current pages for residential or unclassified, I think they are already explained quite well. The content for unclassified in particular is a little imprecise, but I think that is because it is not a precisely defined class. Which is exacly why I started this proposal. Do you have specific enhancements for the existing pages in mind that could solve the issue in another way? Martijn van Exel (talk) 19:58, 14 July 2021 (UTC)
I've always wondered why unclassified is not just called quaternary. This would eliminate any semantic confusion while also consistently emphasizing relative importance.
- @Timmy Tesseract: "Quaternary" is an apt term for this concept. It highlights how we're trying very hard to stick to keywords as opposed to numbers for what is essentially a numeric system. On the other hand, relatively few people know what "tertiary" means (if they didn't study art in school) and fewer people still would know what "quaternary" means (something about dinosaurs?). We can rely on editors to define preset names that are friendlier than the raw tag values, but if the present name is going to be something like "Minor Road", then we might as well make the raw tag value match. – Minh Nguyễn 💬 20:07, 12 July 2021 (UTC)
- When I first read the the hackmd pad about this, I was confused with what problem it was trying to solve. I agree with @Timmy Tesseract: that merging these two road types will lead to information loss. So, the real issue here is that "unclassified" is confusing enough for English speakers (what more for those who aren't?) and I don't think "quaternary" is going to be of much help on that regard. So, plus one for @Minh Nguyen:'s "Minor Road", or highway=minor? GOwin (talk)
The other issue is the loss of the difference. A router should penalise residential roads over unclassified.
On residential roads there is far more likelihood or children playing or riding their bikes. Also residential roads are typically less than 4.5m wide, so a really bad place for two large vehicles to meet. An unclassified road, passing through a predominately residential area needs to be favoured for through traffic, say trucks who need to reach a factory.
Unclassified and residential have very distinct and even mutually exclusive definitions. The first is the lowest level of connecting roads while the latter explicitly does not have that function. Lumping them together would unnecessarily remove important complexity.
In theory some of the differences could be expressed through other tags. But I don't think it's realistic that attributes like width, speed limit and speed bumps will ever be sufficiently surveyed for the 75 million roads that would be affected by such a merge. In addition to that there are further characteristics which cannot be accurately reflected through tags e.g. lower practical speed limits due to lanes being obstructed by parking cars, playing kids, pedestrians etc.
From my observations lack of distinction between unclassified and residential is mostly an issue in areas where significant amounts of roads were added through imports (e.g. tiger) or by remote mappers (mostly corporate mapping teams, to a lesser extent also hotosm, often aided by primitive AI). In places with active local mappers distinction tends to be very nuanced.
To what extent are the existing distinctions between highway=unclassified and highway=residential important enough to data consumers to keep as separate highway=* values, versus relying on secondary tags? This proposal is happening in parallel with an effort to rationalize highway=trunk tagging in the U.S.: the previous practice was tied to physical characteristics that are more effectively expressed via expressway=*. Maybe whatever attributes local communities have been relying on highway=unclassified to express can be expressed in other ways?
It's true that we're unlikely to microsurvey every minor street, but I'm not sure a data consumer can confidently make assumptions about width and speed bumps based on road classification anyways. Routers do make assumptions about speed limits based on road classification, but this is nothing more than a last-resort, low-confidence fallback that is typically biased towards whatever road conditions the developers are familiar with. (highway=unclassified does influence turn channel detection in Valhalla , but I think it's pretty minor.)
I wouldn't necessarily blame imports and remote mappers for inconsistency in unclassified/residential tagging. For example, the local mappers in Ohio have long standardized on tagging township roads as unclassified roads, because they're generally higher-importance than other local roads but not important enough to be tertiary. From our perspective, it was a slot that needed to be filled, so we filled it with the most intuitive distinction we could think of. Certainly it was a more useful distinction than one about width: 4.5 meters (15 ft) is narrower than most driveways and parking aisles in the U.S. suburbs.
- @Minh Nguyen: Proposal author here. I recognize that in some jurisdictions / countries / communities, conventions have emerged that make use of highway=unclassified as a distinct class. While I personally question the need for having that many road classes, I wouldn't want this proposal to 'override' those conventions in any way, or imply that 'they are wrong'. However, I think that for many countries and many mappers, the differences between the lowest road classes are subtle enough that they make mapping harder, not easier. When I started thinking about this, my first thought was 'well, we can just tell mappers to ignore unclassified'. But that is hard to do when editors offer them as presets, and there's many highway=unclassified out there tagged on a wide variety of roads. Perhaps a middle ground could be highway=minor / minor=unclassified and minor=residential? Then we keep the granularity while giving mappers an easier option. Martijn van Exel (talk) 19:49, 14 July 2021 (UTC)
- @Mvexel: I don't know that the mappers in Ohio necessarily care to retain the distinction between highway=unclassified and highway=residential. As I mentioned, the mere existence of an additional classification meant mappers needed to find something to use it for. If that classification goes away, I think we can cope. That said, it would be interesting to consider the impact on the rest of the classification scheme. In the Midwestern U.S., there's a tendency to assign lower classifications than in urban areas along the coasts. With fewer possible classifications, I wonder if we would see a trend of promoting former unclassified roads to tertiary or demoting them to residential, and whether some tertiary roads would be promoted to highway=secondary and some residential roads demoted even further to highway=service in order to maintain a legible distribution of classifications. Whatever the case, I hope we don't adopt such rigid definitions like 4.5 meters (15 ft) that a classification becomes inapplicable to an entire region. – Minh Nguyễn 💬 20:58, 17 July 2021 (UTC)