[OSM-talk] Misclassified roads

David Earl david at frankieandshadow.com
Wed Jul 9 17:17:20 BST 2008

On 09/07/2008 16:03, Steve Hill wrote:
> On Wed, 9 Jul 2008, Chris Hill wrote:
>> I would be strongly against a global change of highway=unclassified - all of 
>> the roads I have tagged as unclassified deserve to be so.  I have been 
>> working partly on a very rural area, where many of the roads are unclassified 
>> (country lanes).  To have to retag them from road to unclassified would be a 
>> very annoying waste of time.
> I agree that there are areas where the classifications are accurate, but 
> is there a good solution to the problem?
> I'm starting the discussion because I think there is a real problem here - 
> I don't have the solution, I'm hoping that a discussion might produce one. 
> :)

We've gone round and round the issue of what road classification means 
many times before. With a few dissenters, the consensus has generally 
been that you tag what you find on the ground. This sometimes 
contradicts the "official" classification. Some people who have had 
access to this information have used different tags to apply the 
"official" classification (though I do wonder about the copyright status 
of such information).

While in the UK (in some other countries it is much less clear cut) this 
contradiction only happens occasionally for trunk (green signs), primary 
(A number on b/w signs) and secondary (B number on b/w signs) - e.g. the 
A road through Oxford discussed a while back - it is pretty much 
universal with the lesser roads which "officially" have a 'C' 
designation but which is virtually never signposted (or commercially 
mapped) as such, and is really a convenient shorthand for highway 
engineers. Even rural footpaths are numbered, usually uniquely per 
parish, but usually only evident when you come across a formal 
diversion/closure notice.

So for the lesser roads, we have what amounts to a subjective choice: 
residential, tertiary, unclassified (and ok, track, service, byway etc, 
but those are perhaps a bit easier to be objective about).

As it is subjective, I think you are wrong to change them except to 
maintain a consistency of approach in an area or where they are just 
wrong (signposted as a B road for example but not tagged as secondary).

What I've done (and what a lot of others also seem to have done), for 
what is now getting to be a very large contiguous area - maybe 2,500 sq. 
km. centered on Cambridge(*) - is
   - tertiary for
    (a) unnumbered roads connecting settlements, except where they are 
so narrow that they really can't be considered as a reasonable 
connection even when they do actually join settlements (and yes, that's 
    (b) unnumbered through or key distributors in urban areas (along 
with abutters=residential)
   - unclassified for
   (c) the rural exceptions to (a)
   (d) with abutters=something, for urban roads which are seriously 
non-residential (e.g. public roads through an industrial estate)
   - residential for everything else public, surfaced and unnumbered in 
an urban area, in which I include possibly only partly residential.

The break between residential and unclassified (or between tertiary with 
and without abutters=residential) is not visible on the renderings, but 
I've felt it is important to leave it in as I think it is a potentially 
useful distinction for e.g. defining an "urban envelope" or applying a 
reduced default speed limit for journey planners.

IMO, this gives a good indication of a hierarchy in rural areas which 
continues in and within urban areas. This leads to a nice rendering, but 
it isn't just tagging for rendering, it genuinely reflects the hierarchy.

OS Landranger maps have a similar approach, based on width less than 4m 
(by memory). While my estimate of width is subjective, not measured, I'm 
essentially doing the same thing (though sometimes I am inclined to make 
a somewhat wider road unclassified if it goes nowhere).



- the area now contiguously complete to the "all streets with names plus 
main POIs" level is mostly rural, covers around 150 villages, 6 market 
towns and one modest city, and extends roughly west as far as Papworth 
Everard, south to Ashwell, Royston, Barley and Saffron Walden, est as 
far as Moulton east of Newmarket and north as far as Littleport north of 
Ely and the Ouse south west of Ely.

More information about the talk mailing list