[OSM-talk] License plan

[Simon Ward]

On Wed, Mar 04, 2009 at 12:33:56AM +0100, MP wrote:
> This could be perhaps "optimized": if user A creates some
> highway=road, user B changes it to residential and user C changes it
> to secondary. A and C agrees to new license, B won't.
> But contribution of B was completely removed by C's edit, so it won't
> be necessary to revert to highway=road in this case. Basically, if the
> edits of "incompatible users" got later reverted or altered so their
> contribution is not there anymore, there is no need to rollback, just
> delete their revision from history.

This seems reasonable, but (there’s always one) what happens in the case
that A creates highway=road, B changes it to highway=residenital
(intentional mis‐spelling), and C corrects it to highway=residential?

Unless C can be said to have surveyed it, this looks like an
“improvement” to B’s efforts, and a trivial one at that.  It should
probably be reverted to A’s edit, and tagged for resurvey.

> > There is the idea floating around that modifications to existing data
> > are insubstantial, and successive contributions could potentially be
> > kept without issue, but I think it is safest to remove them.
> 
> Perhaps for really minor changes, like alterations to created_by or
> conversion from "true" to "yes" or alike we could make an exception.

Reasonable: Changes that don’t change the semantics, or are just
meta‐data about the change, can be excepted.

> Would there be at least some information like "this object was
> reverted because of new license" (which would signal that the object
> perhaps need to be re-improved somehow) and for deleted objects
> information that "something was deleted from here"?

I don’t see why not.

> Also, what if someone who disagrees to new license deletes some data
> (either because that data is wrong or is replaced by something else
> that he draws). Will the deleted data get restored?

I know what OSM needs:  Changesets! ;)

I think all incompatible edits should get restored, although I
understand it could lead to a little bit of a mess.  Hopefully, in most
cases:

 1. A scribbles on the map. [compatible change]
 2. B removes the scribble [incompatible change]; and
 3. B replaces it with a neat road [incompatible change].

B doesn’t agree to the licence and the neat road gets deleted, and the
scribble gets added back in.

The following looks more messy, however:

 1. A scribbles on the map [compatible]
 2. B removes the scribble [incompatible]
 3. C adds a neat road [compatible]

If we follow the rule of reverting incompatible changes only 2 is
reverted to 1 (A’s scribble gets added back in).  3 is considered an
independent change.  We end up with both a scribble and a neat road in
the same area.  This situation likely won’t be easy to detect until
after the changes, when validators will gleefully litter the map with
warnings about overlapping ways.

Simon
-- 
A complex system that works is invariably found to have evolved from a
simple system that works.—John Gall
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20090304/1e41a414/attachment.pgp>