[OSM-dev] likenesses

[Matt Amos]

On Thu, Oct 1, 2009 at 11:33 AM, Jonathan Bennett
<openstreetmap at jonno.cix.co.uk> wrote:
> Andy Allan wrote:
>> We have already, it's been made and engineered and everything. Don't
>> mistake endless unproductive waffle on the mailing lists as evidence
>> that a problem hasn't been solved - it's just people who aren't
>> capable pleading to the gods to come and save them from themselves.
>>
>> http://wiki.openstreetmap.org/wiki/Osmosis/TagTransform
>
> That's a way of dealing with the problem at a data level, but it doesn't
> deal with it at a people level.
>
> To use TagTransform you either have to accept the presets (which could
> be out of date) or come up with your own transformations, which could
> take a lot of research and still not reflect what mappers are doing.
>
> If I understand what Steve is talking about, he wants a way for mappers
> to *document* what they think the equivalences are between tags. So that
> would be the source material you'd plug into TagTransform.
>
> For what it's worth, I think that's the wrong way of going about it. I
> think multiple descriptions of the same tag in different locales is the
> way to go.

+1

there's clearly two places to do this transform; in the editor (i.e:
presets) and as a filter before the app which consumes the data (i.e:
tagtransform). the two places reflect very different types of
transform, though.

translations in the editor are clearly for the benefit of the user,
who might prefer to see straßen-typ=autobahn rather than
highway=motorway. but what really distinguishes an autobahn from a
motorway? i assume there's several differences in the laws regarding
them, but the primary difference is; all the motorways are in the UK
and all the autobahns are in Germany. so it would be inappropriate to
tag anything straßen-typ=autobahn in the UK - even if you prefer to
have the UI in German while you're editing the UK. isn't it easier to
let the editor handle language and locale-specific presentation of the
tag, but leave the data more-or-less consistent?

translations in the consumer are there (usually) to reduce the
vocabulary of OSM down to things that the app can handle. a simple
example might be that, for all their differences, it makes some sense
to render residential and unclassified roads as the same "minor road"
type. this, also, is probably locale-dependent. there have been
discussions on talk@ before about whether footpaths are also
permissive cycleways in various countries and what the default
implication of higway=footway should be. (disclaimer: i only dimly
remember this, so i may have got it totally wrong). again, it makes
sense to have a fairly universal set of tags in the data with a
locale-specific translation. this translation will also be
domain-specific; only a cycle-specific renderer, or a cycle routing
engine would care about the distinction in the previous example.

while i'm all in favour of a library of tag transforms and editor
presets, i'm against having any fixed, defined set of tag
equivalences. that seems to me like the beginnings of an ontology.

cheers,

matt