[Osmf-talk] Tagging standards

[Peter Gervai]

On Sun, 16 Oct 2022 22:33:12 -0400
john whelan <jwhelan0112 at gmail.com> wrote:

> I strongly suspect assets are tagged differently depending when they
> were mapped and even who mapped them.

I'd like to mention a bit different aspect to the one already discussed
here:

It is often a problem when a specific tag is used for either different
purposes or even worse when it's used for the same purpose but the
values are interpreted differently.

While machines can be taught to normalise keys and take "traffic_light"
and "stop_light" and convert them internally to "traffic_signal", they
cannot be taught that "smoothness=bad" in Ungaristan is the same as
"smoothness=horrible" in Chillandia for a general case. Same often
exists by track grades, or paths vs. tracks (don't ask me why), or
even "highway=service" vs. "highway=unclassified" or various attributes
I've seen being too loosely defined (in the past at least) that their
use is extremely hard to normalise to a known value by code. They can
be looked at by a human and that's all they're good for, and mere humans
look at tags rarely nowadays.

Wiki got much better in recent years, specifying and giving examples
for values; and I am absolutely not against subjective values until
the users are somehow universally guided about the expectation of the
results. Without the wiki people would (and indeed do) use wildly
inconsistent values on the border of useless.

So, I support documenting tag usage as detailed as possible to make it
feasible to actually use them in software, before people start using it
based on wildly different bases. Not compulsory, but definitely advisory
kind.

g