[OSM-talk] Map features discussion & voting

[David Earl]

> -----Original Message-----
> From: talk-bounces at openstreetmap.org
> [mailto:talk-bounces at openstreetmap.org]On Behalf Of Frederik Ramm
> Sent: 11 June 2007 09:54
> To: Tom Chance
> Cc: talk at openstreetmap.org
> Subject: Re: [OSM-talk] Map features discussion & voting
>
>
> Hi,
>
> > Unfortunately I think the process is massively failing right now.
> > Lots of the emails Alex has sent indicate decisions on the basis of
> > three or four "votes" - three or four people in a mapping community
> > of over 700, on the basis of very little discussion. I really think
> > there ought to be a bare minimum, like 15 votes with a 75%
> > majority, before something can be accepted.
>
> Depends on what these votes actually mean. Since anyone is free to
> tag whatever he or she wants, and anyone is free to write or modify
> renderers to draw exactly what he or she wants, the "Mapping
> Features" page is not all that important. It offers guidance to new
> mappers, but if someone wants to use "village_green" then by all
> means he can do it.
>
> These votes are not about what is allowed or forbidden in mapping,
> nor about what will be rendered and what not. If we cannot reach a
> consensus on some items, then this only means that we have a number
> of mappers dealing differently with the same thing. Which complicates
> matters, but if it makes the community happy, then that's paramount.
>
> At some point we will probably have to switch over to a system with a
> fixed set of allowable tags, or object types, or whatever, but then
> we'll also have to modify editors and the API to make sure that
> everything in our data set is "correct" as  per these rules. We do
> not have such a strict system currently, so frankly I view the voting
> process (and especially the discussions around them) as more of a
> brainstorming than a decision-making thing.


Oh dear.

I think the tag system is excellent for three reasons: they allow new things
to be introduced without a technical change, and they allow specialist
information to be added where wanted.

But their flexibility is their Achilles heel too. If they are used in a
non-consistent way, we end up with an anarchic mess that has no hope of
being a unified map; as the amount of data grows, the possibilities of
bringing aberrant features into line gets harder and so they'll just get
lost.

What holds it together at the moment is (a) the Map_features page, as
guidance and (b) (IMO far more influential) what is actually rendered on the
public maps.

While in theory anyone can draw what they like, in practice this isn't the
case. For specialisms, yes, but adapting your own copy of osmarender, for
example, doesn't get your feature in the mainstream. If it appears on mapnik
and/or osmarender centrally, most people will go with the flow, and that's
what's kept things sane IMO.

Take, for example, the "anarchic" tagging place=small_town. There are six
places marked like this (Totton, New Alresford, Grange-over-Sands, Arundel,
Hythe and Markyate). What has happened to these is that all but Totton have
had additional place=town or place=village nodes added nearby, probably
because no one but the author knew about the existing nodes. Totton, the one
that doesn't have an alternative, doesn't appear on mapnik or osmarender at
all because of its non-standard designation. That's not very helpful for
those who want to make use of the map, which ought, in time, to be many more
than those who contribute.

(Maybe there is a need for something between a "village" and a "town" - I
would agree: yesterday I tagged as village one settlement with 150 people
and one with 4,500 which is quite different - but just inventing one on the
fly doesn't help because no one knows about it and the information gets
lost.)

The exception is if you happen to be one of the people programming the
renderers. If something is invented *and* gets rendered, it is likely to be
adopted, vote or no vote. waterway=stream is a good example; there are many
others.

If all the consumers of the data (renderers, route mappers, indexes etc)
understand that e.g. natural=wood and place=wood mean the same thing (a
recent example); or that waterway=stream and a combination of waterway=river
and width or estimated_width=3m mean the same thing, we *can* cope with
variation, so long as it is *known* variation. There's possibly good reasons
for the latter, but the former is just gratuitous and unnecessary. Not
finding tag names aesthetically pleasing is a pretty poor reason to
introduce random variations: the effect is to either throw your own work
away, or to impose on others the time involved in dealing with pet
variations.

Whether standardisation is arrived at by iteration, democracy, or the whim
of map renderers may not matter that much. I think the map renderers in
practice have the whip hand: village_green for example scraped through the
vote, but I think for it to become useful or truly adopted I would have to
spend time changing the rendering myself or persuading others to render
include it; if it were rendered already others would use the tag too.

So, it may not be "forbidden" to use non-standard tags, but it is pretty
pointless and limits the collective potential of real use of the map in the
future. Map_features, far from being "not all that important", is vital to
producing a usable, useful end product.

Having said all that, I agree that the voting system doesn't work well; and
it's disappointing that most of the  discussion is often around the
aesthetics of tag names rather than the real concepts behind them.

David