I tend to agree that for some roads the classification simply by digits does not work well.
It does work probably well in telling the “classification”, but in practical life we want the highways rendered prominently on a map based on importance. For the similar reason we want them to be preferred when routing on OSM data.
Right now, for a lot of data consumers (mainly map rendering or routing), the road classification gives them some reasonable defaults.
You can probably assume that a highway=secondary is paved. Or that a highway=primary has better driving conditions than a tertiary (like number of lanes, overpass instead of traffic lights or such).
We have to maintain a robust set of critera on how we can tag and make this tagging verifiable by other mappers. It is like in science. If you claim something there must be a possibility to prove you right or wrong.
So there must be criteria that if I claim a road is a primary, then others on the ground must be able to verify that. And in case road-work happened maybe a road can be re-classified based on this criteria.
The classification by digits served well for an initial classification.
And I think we discussed it before that there might be defined criteria which allow “upgrades” of roads in a higher category. A four-lane road with shoulder and a middle-divider might be upgraded from tertiary into a secondary.
Maybe we can also come up with some criteria which allows to downgrade roads. Or hard criteria which will prevent a specific classification.
I for example would not want to see any unpaved road in Thailand with a classification higher than tertiary.
This is a criteria which can be easily verified by others on the ground.
Another criteria which is verifiable might be the intensity of usage of a road.
Taking the example from Russ: So technically (based of DRR database) a section of a road might be tertiary. But based on the intensity of usage, it could still be used as a primary road. Number of vehicles per hour might be a criteria.
By counting vehicles on the ground for half an hour you should be able to get a good idea how “intense” a road is used.
So in this case it would “protect” a road section from a downgrade to create a more uniform grid of the road network.
I remember also a tertiary road (according to DRR database) in greater Bangkok which was just an agriculture track. Maybe planned and earmarked already for future expansion, but not yet built.
For such cases I would like to have some objective criteria which allows to not tag it as tertiary.
The classification should give a hint, but as government databases can be wrong as well, the authority should be the ground-truth by mappers.
But not based on “opinion, impression” or other subjective criteria. I saw edit-wars where people wanted to do poor-man’s traffic management and reduce via-traffic through their quarter by messing around with the database in the hope that routing engines would prefer other ways.
So what kind of “minimum” requirements would we have for each specific road type (unclassified, tertiary, secondary, primary, trunk)? surface, width, number of lanes, other factors? Are there “de-railers” which prevent a category? What could allow up/downgrades? What could allow a road to keep its classification a bit longer?