[Tagging] Thoughts on how to replace or modify an exist/established tag (Was: Feature Proposal - RFC - sluice_gate)

[Simone Saviolo]

2011/1/6 Ralf Kleineisel <ralf at kleineisel.de>

> On 01/06/2011 06:00 PM, Peter Wendorff wrote:
>
> > What beside of this - I fear, stupid - "certification" is the benefit
> > for a hiking map in supporting e.g. maxspeed of motorways as part of the
> > OSM core being the decision basis to get the certification?
> >
> > To make a better example: Garmin AiO for Europe is getting too large for
> > many devices currently - so the core definition you propose would
> > require to include buildings in the map, no matter of their size and the
> > drawbacks of excluding most old devices by including the building layer?
>
> I think it's even impossible to support all OSM tags on a specific map.
> When rendering a map or converting the data into a special format (e.g.
> Garmin devices) you always have lot of constraints.
>
> For example Garmins don't support an unlimited number of way types. So
> you have to map OSM tags to device tags, which is not always a lossless
> process.
>
> In addition it's often not even desirable to show everything - many maps
> have a special purpose and should only show the things relevant for that
> purpose.
>
> I think that map renderers and map makers already do their best to
> support as many useful (for their map!) tags as possible and most
> mappers do their best to map according to the wiki definition because
> they want their work to appear on the maps.
>

Yes, and I think of this "certification" (it may not be the best name for
it) as a way to indicate that a certain application supports all the map
features from a set that has been considered essential.

To reply to Peter, I would think of a "core" set of features (including main
geographic features like coastlines, rivers or country boundaries, and a
minimal set of other features, such as highway=* or place=*) that should be
included in any map; on top of that, other more specific features would be
in a specialized set, for example "Cycle Map Features" including the main
tags related to cycleways. Thus, a map for cyclists that has both the Core
and the Cycle Map Features should be considered at least a decent map - in
that it correctly handles the information that is useful for its intended
users. A map for boat navigation of a lake would of course not need the
Cycle Map Features "certification".


> Remember, the goal of OSM is a FREE map, not one with a lot of rules and
> restrictions. There are enough restricted maps out there.
>

Sure, but it's a matter of defining a line between "chaos" and "coordinated
anarchy". It seems that many OSMers would not want to go any further than an
anarchy, this may be ok, but for the data to be somehow useful (and not only
a bunch of self-referential, self-glorifying material) coordination is
needed.

We need to know that oneway=-1 is either "ok" or "not ok", so that consumers
know whether they should handle it or not. If you lean on the side of
supporting anything you see in the database more than X times, then you get
very complicated (and costly!) consumers; if you lean on the side of using
only what's documented in the wiki, you risk leaving out a lot because some
mapper in his basement decided that to him it's ethical to use oneway=-1
rather than reverting the way's orientation.

This step is needed if we ever want to do something serious with the data. I
visited a few places these holidays, and wanted to check the map there for
errors or updates. To make a very simple example, I found several
roundabouts not tagged with junction=roundabout. Some were circular ways
with oneway=yes (some had oneway=true), some had area=yes on them, some had
the importance of the most important incident road, others didn't. Now, in
this case it was easy for me to fix them, because roundabouts have a
well-agreed tagging scheme, but if the feature had been one of those for
which there's controversy, what should I have done? I couldn't have done any
QA, because "if the mapper marked that like this, then I should probably
respect his decision" - otherwise I may start an edit war.Or was it a poor
edit, maybe by a newbie, that never got fixed - how would I know?

We're at the point where local groups of mappers map the same thing their
own "regional" way, because Germans are for x=y while Italians think x=z is
better, and Italians from Lombardy are fans of x=z2, but Americans think the
feature is not even relevant enough to be mapped. There's never going to be
any useful map larger than a few hundred kilometres this way, except for a
few "mature" features.

I hope this opinion doesn't get flamed as "anti-wiki". I'm all for wikis,
but a minimum of control is needed. When Wikiversity was created, I
suggested that some users with proven credentials be put in charge of their
own matter: for example, that one or more doctors in Civil Engineering took
care of the Civil Engineering course. I was told that that's not how a wiki
works and that a "wiki"versity should be open to anybody's edits - I left
the project because then the Wikipedia effect is always latent: anyone could
put any garbage in it and it may go unnoticed. I wonder how long it would be
before someone noticed if I added a fictional city somewhere. Maybe there
are no mappers in the area and the city would stay. So people would load the
map in their GPS and be guided through it, only to find rice fields where
they should go.

In other words: if we ever want a map created from OSM data to be used in
anything bigger than a proof of concept, or if we - practically - wish to
see hand-held devices with an OSM map on them someday, then we definitely
need *at least* to be able to do QA, and this can only be done if there's a
set of rules. Of course these rules should only cover a mature subset of the
current tags! But at least those should have such a thing.

Best regards,

Simone
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20110106/7db2c934/attachment-0001.html>