[Tagging] More human readable values for traffic signs

[Richard Fairhurst]

On 29/10/2015 21:52, Colin Smale wrote:
> I don't have any examples to counter your statement. But I am assuming
> you are referring to the use of a spatial database.  It is IMHO a high
> barrier to entry. Are we to expect users to have that kind of
> infrastructure and skills at their disposal? What about mkgmap and the
> many other consumers which simply work with a snapshot in PBF or XML and
> need to be able to do the right thing with the data with lets say a dual
> core and 8GB.

Indeed. It's very possible to "do the right thing" by running spatial 
queries on the data, without a spatial database, working directly on a 
PBF snapshot, and running lightning fast in just a few GB of memory and 
on a desktop-class machine.

I can confirm this because I've done exactly that, in a tool which I'm 
delighted to see is proving popular: 
https://github.com/systemed/tilemaker, and in particular, 
https://github.com/systemed/tilemaker/blob/master/CONFIGURATION.md#lua-spatial-queries 
.

That aside, even if you accept that the interests of the data consumer 
are paramount - which, as a data consumer, I don't, and I'm rather tired 
of tagfiddlers without development knowledge second-guessing what 
developers might need - the number of consumers to whom it's important 
to differentiate (say) UK and French no entry signs is an 0.001% edge 
case, not one worthy of defining the entire tagging model.

You could just as well argue that you're penalising the little guy by 
preventing him from searching simply for "traffic_sign=no_entry" and 
making him search either for 193 country-specific values, or run one 
full-text query. Given your concern about doing things "very cheaply", 
I'm pretty sure more people have a spatial index on OSM data than have a 
full-text index.

But let's remind ourselves of Mateusz's original posting:

> I recently started tagging traffic signs and I am surprised by wide
> usage country-specific traffic sign codes.
>
> I think that at least common signs may be tagged by human-readable
> values.

Quite.

Richard