Hi! In this changeset you changed way/315601384 to be highway=unclassified.

Significant parts of this way are completely unpaved and little more than a mountain track. In Britain, we would expect highway=unclassified to be paved unless otherwise stated (and even that is very rare).

I've changed it back. In general you'll find that most road classifications in Britain are correct so I'd suggest your team shouldn't change them without checking with the local community.

cheers
Richard

To add with reference to this particular example:

There is very very clearly a difference between way/129297661 (which already had tracktype=grade2) and way/370062638 (to which Modest7 has edited tracktype=grade4 in this changeset). If you can't see that from imagery I would strongly recommend you consult your local optician and/or computer monitor servicing technician.

This edit reflects that difference. So it has materially improved the OSM database in this area. Whether tracktype= is the ideal tag is another question (and that is more a failing of an imperfectly conceived, rather subjective tag tbh), but this edit adds useful information so should not be reverted.

A wholesale revert would be an over-reaction.

OSM is iterative - it doesn't have to be 100% accurate at the first attempt, though it should take care never to be actively misleading. Lots of OSM data, particularly in remote or less travelled areas, has been added by reference to imagery. That includes surface quality information - for example, in areas of the rural US without significant tree cover, it's very often possible to discern approximate surface quality from imagery, and doing so is a great improvement on the base TIGER import.

So what Modest7 has done here is certainly not against standard mapping practice in OSM.

Rather, I think the main question is whether data consumers are likely to infer greater certainty from the tracktype= tag than armchairing can provide. There is obviously a difference between "I have surveyed this and it's definitely grade5" vs "I have looked at this from imagery and it seems grade5-ish". (That said, parsing tracktype= is very problematic anyway because international practice varies so greatly: in particular, the values are used differently in France, Germany, and the US, and I have different rules for each in cycle.travel.)

So my suggestion would be that any bulk edit should not be to revert this useful work, but to add "source:tracktype=imagery"; and that Modest7 should add that tag for future edits. (To the affected way, not the changeset; no-one parses way metadata with reference to changeset tags and nor should they be expected to.)

Richard (writing with my cycle.travel hat on)

Ultimately, common tag documentation needs to move to a site with a pull request model - in much the same way as install documentation moved to switch2osm because people kept fouling up the instructions on the wiki. But that would require someone to do the work of setting up such a site.

*take note (not "not"!)

Hi - in this changeset you changed highway=bridleway to highway=service, which is fine, but by adding access=private you've broken bike/foot access. Bridleways in the UK permit bike and foot access. I've reinstated it in changeset/87908293 but you might want to take not for future edits and those by your colleagues.

The intention is to move to vector tiles in due course which will enable client-side language changes, and there's a bunch of work being done on it right now (https://github.com/pnorman/openstreetmap-cartographic). But it's not going to happen overnight.

Fair point on the recent slash addition - sorry, I had thought it'd been done earlier!

(Or put another way, name= should be Welsh in areas where people listen to Bob Delyn and English in areas where people prefer Goldie Lookin' Chain)

Indeed they can. That, however, does not preclude that Welsh names can be used in name= as well. Nowhere on that page does it say "only English may be used for name=".

Long-established international practice in OSM is that the majority language is in the name= tag. So I'd expect to see (for example) English in Monmouthshire and the Pembrokeshire coast, and Welsh on the Llyn.

Where there isn't a clear majority language, local practice is often to use both languages separated by a slash (though it's a bit of a hack tbh). That seems appropriate for Snowdonia.

As a general rule, if you're changing the name of a bloody large feature 15 years into the life of OSM, you should maybe reflect that you're unlikely to have suddenly thought of something that has escaped thousands of mappers for the last 15 years ;)

Hi - you appear to have turned this into a three-mile bridge which I'm guessing was not intentional ;)

Ha - comments crossed!

Thanks!

(I suspect the "road" that's shown on satnavs might actually be the one a short way to the west via Foxholes Farm, which has the width of a road but a very rough surface, and the actual right of way status is unclear.)

Hi. In this changeset you've changed way/35028249#map=16/51.8841/-1.6228 from a bridleway to a path. I was there yesterday and it is definitely signed as a bridleway. Could you shed any light?

You're misunderstanding.

Tags do not cascade down from relations to ways. That isn't how OSM works. Relations are distinct entities that describe themselves.

Your relation is a route relation. It describes a route that follows some ways. It does not describe those ways themselves.

Access is specified by tags on ways.

I'm hesitant to do the whole "appeal to authority" thing, but as the third longest-standing contributor to OSM still active (since November 2004) and the developer of the site that probably does more in-depth tag parsing than any other (cycle.travel), I honestly do know what I'm talking about here. Thank you.

Hi Bill - basically if you tag something as "highway=path" as you did, you're just saying "this is a path". You're not giving any information about who can use it.

You need to add bicycle=yes and foot=yes on the way to make it clear. Route relations are additional information; they don't obviate the need to have the correct way tagging.

In the route relations in particular, note that the semicolon-separated style you've used ("route=hiking;bicycle;foot") is idiosyncratic and undocumented, and most renderers/routers will not understand it.

Thanks for all your work! :)

(Sorry - should add that the same is true of retagging as highway=unclassified - in developed countries this generally means a paved road, so adding highway=unpaved/gravel/dirt/whatever if it's unpaved is really helpful. Obviously lots of unpaved roads round this area :) )

Hi,

Great to see the cleanup work you've been doing!

If you remove tiger:reviewed from an unpaved road, could you make sure to add a surface tag? Otherwise it's impossible to tell whether the road is paved or unpaved, and that breaks bike routing among other things.

cheers
Richard

Hi - by changing this from highway=cycleway to highway=path, you've destroyed the information that bikes are allowed to use this path. I've repaired those that I can see by adding bicycle=yes, but could you check all those you edited? Thank you.

Thanks.

This is what's called a "modal filter" - where a road has been stopped up to through motor traffic, but is still available for bikes and pedestrians.

In cases like this in urban areas of the UK, you should always default to highway=cycleway rather than highway=footway.

I've fixed this instance but would appreciate it if you could review your previous edits, and use highway=cycleway in the future. Thank you!

Hi - why have you changed this into a highway=footway rather than a highway=cycleway?