[talk-au] ODBL yet again, but from a pragmatic approach...

John Smith deltafoxtrot256 at gmail.com
Sat Jul 10 00:18:58 BST 2010


As people should now be aware there is currently there is an issue,
not so much with ODBL, but the new Terms and Conditions people have to
agree to stating that OSM can change to other "free" licenses in
future without requiring consent, while in theory this is a great idea
since if there is a compelling reason to change/upgrade the license
they can do so without all the problems occurring now, however due to
the absence of requiring such a free license to be cc-by compatible
(require some form of attribution) this then means any cc-by data
would now have to be expunged from the system.

Currently we have a fair bit of cc-by data in the system, things like
ABS boundaries and in turn any data derived from such data, but so far
there is only assumptions on how much data is this exactly, especially
in Europe where the assumption is the majority of data has been
relicensed or is clean to begin with, so they don't care about anyone
else who may be effected by this change, but of course the big unknown
is how many contributors will actually agree to this change,
especially some of the more prolific editors.

The $20mill dollar question however is this, and this is the pragmatic
part, what would the state of the map be tomorrow if the license
change over happened if all the cc-by data and derived data
disappeared.

For the purposes of this exercise I'll just make the blind assumption
that anything with attribution=* would be considered cc-by, obviously
this isn't a perfect test since some people have stripped the
attribution information and other data may not have been attributed
properly, then again even ODBL data could be tainted, and subtly
enough to corrupt large chunks of the database, however this should
give us a pretty good idea of what we're dealing with rather than keep
making blind assumptions.

I found that there is 97,573 ways/nodes/relations within an Australia
bounding box with an attribution tag, although there needs to be a lot
more interrogation of the data to make this a much more tangible and
suitable for making objective decisions based on it. Although I did
create a noattribution navit[1] file and a gosmore file[2] to try and
help with visualising.

The above 98k objects make up about 8M of compressed data[3], while
this wouldn't be completely devastating, we're not just talking ABS
data, there is a lot more to it like points of interest and national
parks and other such things.

As Kai wrote in another thread, the loss of data could have a big
demoralising effect on anyone that spent time cleaning up or otherwise
manipulating that data. Those that are so gung-ho to push through
their own agendas might want to push for a small change to the TCs
ensure attribution and most of this discussion would disappear, rather
than alienating[4] people that contribute data from regional areas
that we have enough trouble sourcing by any other means, that is
unless they want to come and recruit others that would also do the
work for free instead.

Although I'm not sure what the point is of moving to another
attribution/share-a-like license, if the TCs undermine this, unless of
course the intent is to eventually force everyone to go to PD long
term, but doing it on the sly hoping no one notices where things are
headed.

[1] http://map-data.bigtincan.com/data/australia-noattribution.navit.bin
[2] http://map-data.bigtincan.com/data/australia-noattribution.pak
[3] 149,017,722 v 157,576,420 respectively
[4] http://lists.openstreetmap.org/pipermail/legal-talk/2010-July/003441.html




More information about the Talk-au mailing list