[OSM-legal-talk] Infringements - examples, analysis and request for removal

Darko Sokolić Darko.Sokolic at xnet.hr
Sat Mar 31 12:25:30 BST 2012



Dear colleagues,

I contributed to OpenStreetMap under CC-BY-SA 2.0 license. It was great 
pleasure, and I enjoyed it very much.
I did not accept new Contributor Terms and new license.
Also, I did not authorise anyone, in any way, to relicense or sublicense 
my contributions.

I expect that my contributions will be deleted as part of moving to new 
licence.
To my surprise, I see that most of my original contributions are already 
deleted and replaced with no noticeable difference.
Looking at the details I saw that just the user attribute has changed 
(often Janjko and SilverSpace), timestamp is diferent, coordinates are 
slightly offset, ID is of course different, and there is no history data.
Replacement data is therefore in CC-BY-SA terminology Derivative Work 
based on my original Work.

This clearly infringes CC-BY-SA in at least two ways:
* by erasing history and replacing author name this violates attribution 
requirement,
* by submitting such Dertivative Work under new Contributor Terms 
attempt is made to license Derivative work under ODbl and DbCL.

This infringing data must be removed.
Alternatively, while OSM is still served to public as CC-BY-SA, this 
infringing data might be reverted to original data.

Though I like revering data better, I cannot do it. Revertion scripts, 
in my understaning, run on same API for submitting new/edited data. So, 
if to run and of these, I need an active account, an account that 
accepted new Contributor Terms. By doing so, all data (re)created 
through revert scripts will be licensed under ODbl/DbCL. So, I cannot do 
it this way.

I could do removal of infringing data, but this might appear like 
massive vandalism.

Can OSMF revert infiringing changesets, or remove infringing data? This 
is maybe the best way, and also it will probably use server resurces in 
most efficient way. I also believe that OSMF has already tools to find 
similarities in present and historic (deleted) data.

I am not aware of such tools, so I did some analysis myself, developing 
needed tools. Scope of analysis is limited to contributions of three 
users (myself, and two mentioned above, that I noticed by looking at 
live map tiles).

I this analysis I've covered 7329 nodes.
I was looking for situations where any of these nodes is deleted by 
another user, and then new node is created on similar location in the 
same changeset. Then I grouped results by positional error, that is 
distance between new and old node.

This is what I found (grouped by author of replacement nodes):
for positional error of up to cca 11 m in latitude and 7,8 m in 
longitude (that is 4 decimal digits in LAT/LON in OSM database):
  SilverSpace |  4565 nodes (62% of all analysed nodes)
  Janjko      |  1363 nodes (19%)
for positional error of up to cca 1,1 m in latitude to 0,78 m in longitude:
  SilverSpace |  2909 (40%)
  Janjko      |   758 (10%)

For first group we might argue that cca 10 meters is large distance and 
that any usual remapping would fit in (but visual comparison of rendered 
data reveals similarities).
In the second group, where positional error is up to cca 1 m - it is 
very hard to defend this as not infringement.

I started to analyse not only maximum deviations, but averages, and 
standard deviations, and also I looked into minimal positional errors. 
And the I found that significant number of replacement nodes are placed 
on the _very_same_position_ of original node (again - in the same 
changeset):
  SilverSpace |  2235 (30%!)
  Janjko      |   260 (3,5%)
We are talking here about precision of lat/lon in 7 decimal places. This 
is precision of about 11 mm in latitude and 7,8 mm in longitude. In 34% 
of sampled data. This is not a coincidence. This is intentional 
infringement.

If anybody else suspects that his/here data is infringed in similar 
fashion, I am willing to share my tools and experience that I've gained 
during this analysis.
I also indend to refine tools to cover more similaritites. So far I 
dealt only with nodes, their position, and with changesets in which 
nodes were created and deleted.

DarkoS



More information about the legal-talk mailing list