highway=unclassified vs highway=track

I think some of our tagging conventions stray from the global conventions for no good reason

Our conventions:

  • highway=track: Any unpaved road regardless of routing importance.

  • **highway=unclassified:**Smaller paved roads in non-residential places. e.g. agricultural roads, forest roads, roads from behind one village to behind another village, etc.

Global conventions

  • highway=unclassified: is a bit more important in the road grid - it interconnects towns/cities/villages. It is assumed to be paved by default, but it is not necessarily so.

  • highway=track: is mostly agricultural use, forest tracks, usually unpaved but necessarily so.

Like all roads, pavement status can be explicitly indicated with the surface tag in both cases.

This difference may create practical problems:

My only question regarding your post is what about service road - is it “between” unclassified and track or above unclassified in the road grid?
I’m not sure I fully have a grasp of the difference.
In any case, I don’t see a good reason to stray from the global conversions, but an effort should be made to fix this if you want to align all usages…

For decades, before OSM or the internet were created, the map conventions in Israel make a clear distinctions between a paved (“דרך סלולה”) and an unpaved road (“דרך עפר”). That’s a very good reason to set local guidelines.

going back to the British Mandate era:

In fact, the general guideline for unclassified roads declared itself as secondary to country-specific and even region-specific guidelines:

On the other hand, the “practical problem” description seems to me like a classical example of mapping for the renderer. It looks like you don’t like how Mapnik is rendering the map in your vicinity.

Service roads are for… well, providing specific services: A short road for a parking lot, house, or a fuel station entrance, a fast-food drive-through road, etc. It is less important than both track and unclassified when it comes to routing, it does not connect different places, and it’s usually very short.

Hi zstadler,

Mapping for the renderer means tagging incorrectly to make things look pretty. On the other hand, I am trying to adhere to the global convention that the “highway” tag is about routing importance, and not about paving. Yes, it probably does make rendering prettier, and it probably does make routing better. Those are very good side effects.

To take an extreme example: You wouldn’t see a country using the highway tag to indicate the lit status of roads or whether the road is a fee road or not. That place-specific leeway is not to be interpreted too liberally. The highway tag is about road importance, and as long as one is adhering to that, country-specific or region-specific variations are ok. But we’ve ditched the importance difference between “track” and “unclassified” and are tagging something else entirely.

I am pretty sure this is not Israel-specific, and that there are many worldwide non-osm maps that tag by “dirt” and “paved”. I am not sure how this should affect osm tagging.

Please note that renderers probably still make a color/thickness/rendering distinction when “unclassified/track” is combined with “surface”, and if they don’t, then perhaps they should be fixed, not the tags (otherwise this would be mapping for the renderer). So this change proposal is not necessarily proposing to affect Israeli/IHM legends.

Edit: The issue was first mentioned here https://www.openstreetmap.org/changeset/31346575 by dsh4

It seems like this discussion is not productive.

I was quoting the wiki page for highway=unclassified, not changing the meaning of the highway tag to its lit status.

I am sorry if it was understood that way. Perhaps the “talking tone” is lost in messages. I am trying to make a constructive and objective point and am willing to stand corrected :slight_smile: . I tried illustrating with an obviously absurd example but perhaps that made my point vague. Let us review the quote again:

I am trying to say that the above quote above does not allow arbitrary regional guidelines. It is talking about regional differences in the definition of the “level of importance”. For instance, this is why it’s OK for Israel to tag primary/secondary/etc based on sign color. Other quotes state very explicitly that track roads can be paved, and unclassified roads can be unpaved. It is also clear that the highway tag is about the level of importance, and not about other attributes such as pavements (or, more absurdly, lit status).

Yes, it’s a very good reason to set guidelines for renderers (but not for taggers). You are welcome to propose Israeli guidelines that say that “paved roads should be rendered differently to unpaved roads”. Renderers that target Israeli users would follow those guidelines (even when generating maps of, say, Nepal targeted for Israeli users) and renderers that target other demographic groups would not (say, a map of Israel rendered for Nepali users would not follow those guidelines, but the Nepali customs).

The value of the “highway” tag reflects a road’s importance in the regional roads network. That does not change from country to country; OSM has, by design, only one data model for the whole world. According to that data model, how roads would be rendered has nothing to do with how they should be tagged.

Your use-case would still be served: if you want a map to make a distinction between paved and unpaved roads, you should write a renderer that consults the surface=(paved|unpaved) tag when choosing how to render a road. When the “surface” is not tagged explicitly, that is when region-to-region variation is allowed: “surface=unclassified” in England should be assumed to be paved and “surface=unclassified” in certain parts of Africa should be assumed to be unpaved.

Choosing between “track” and “unclassified” to match the rendering convention (the one in the pictures your posted, or any other) would be tagging for the renderer.

I agree, zstadler: the problem is not that Mapnik renders highway=track as less important than highway=unclassified.

The problem is that all consumers of OSM data consider highway=track as more minor than highway=unclassified; Mapnik is simply a proxy (and a bit of a special case since it’s designed with mappers as its target audience).


Regional variation is also allowed in defining “importance”; A small, narrow, infrequently used road road could be “secondary” or “primary” in Africa if it’s the best one in the area, and no better / larger / more important / more frequently used roads are available in that particular region. The same road would have been “track” or “unclassified” if it were somewhere else. Example: https://www.openstreetmap.org/way/441102489

@zstadler I’d like to continue this discussion and hopefully reach some common ground. Would you like to comment on my reply to your (quoted) point? I look forward to your response.

After some long thoughts…

There are two separate issues under discussion here: paved tracks and unpaved unclassified roads.

I tend to agree that tracks can have paved segments. IMO, effectively there is a continuum for surface conditions that allow 2WD cars on tracks: from gravel tracks (e.g., שבילי קק"ל), through deteriorated paved roads, to short segments of concrete on steep sloped of forest roads. I agree that such roads can be classified as highway=track with tracktype=grade2 and one of the paved surface types.

I find that tagging unpaved roads as highway=unclassified, in a small developed country such as Israel, is inappropriate and would be misleading to both local map reader and visitors. This is clearly not the case for Gaza strip and the West Bank which have plenty of unpaved unclassified roads, unpaved residential roads, and even unpaved tertiary roads.

Indeed, the global guidelines for highway classification take the size of a country/area and their development level into consideration. This is true not only for unclassified roads. The wiki for highway=secondary explicitly says “In developed countries it normally has 2 lanes and the traffic for both directions is usually separated by a central line on the road. In areas with worse infrastructure road quality may be far worse”. Developing countries, including Nepal, have unpaved primary highways. A visitor from Nepal would not expect the same standard to hold in Israel.

About tracks: I agree.

About unclassified: I agree that the physical characteristics of highway=unclassified vary by country and that in Israel highway=unclassified should imply surface=paved. In Israel, being unpaved should be a prima facie indication of low importance: if a road is unpaved then it’s probably too minor to be highway=unclassified.


Here are two more examples:

https://www.openstreetmap.org/way/330060970 (Shalva / Ahuzzam)
https://www.openstreetmap.org/way/267931202 (Bnei Zion / Gaash)