[Rebuild] Switching a tile server over to clean data

[Paul Norman]

> From: Frederik Ramm [mailto:frederik at remote.org]
> Subject: Re: [Rebuild] Switching a tile server over to clean data
> 
> Hi,
> 
> On 04/17/2012 08:22 AM, Paul Norman wrote:
> > I'm also assuming that the attribution of CC by-sa tiles with CC by-sa
> > data and CC by-sa tiles with ODbL data will differ.
> 
> Yes although I could see some leniency in the period immediately after
> the changeover, i.e. I think OSMF should say "for a few months after the
> changeover, even if it is technically an attribution requirement
> violation if you just leave the old "Tiles CC-By-SA Data(c) OSM" in
> place, we'll not prosecute you for that".

That would be good. I'm not sure that OSMF can make it binding on
contributors but I doubt anyone would seriously want to sue or send takedown
notices over that.

> > If they were to render tiles from this database you would be mixing CC
> > by-sa only data (the coastlines) with ODbL only data (the new planet
> > file). As these are incompatible licenses, they could not do this.
> 
> No, that is not correct. Since the two databases (shapefile with CC-BY-
> SA coastline, postgresql database with ODbL content) are at no point
> mixed into a common derived database, you can make produced works and
> distribute them.

I see a tile image falling under a derivative work, not a collective work,
as defined in CC BY-SA 2.0. Unlike a periodical issue or anthology it is not
possible to distinguish the different sources. CC BY-SA 3.0 is clearer on
this.

Also - I'm not sure why it matters if it's a database where they're mixed or
an image where they're mixed - they're both works protected by copyright.
3.0 is clearer on this point

> The proper attribution would have to explain that the tiles are CC-BY-SA
> and contain CC-BY-SA OSM content as well as ODbL OSM content, but there
> is nothing impossible about creating tiles that have ODbL as well as CC-
> BY-SA (or even non-open data sources).

You could mix ODbL and non-open data sources in a derivative work, but not
CC BY-SA and non-open sources. 

In any cases, the coastline is not looking like a major issue, we're down to
50 breakage points. I don't really want to get into a lengthy discussion on
this point since it doesn't impact the possible solutions for the caching
issue.

> After you have updated your coastline to ODbL, you will even have three
> different sorts of tiles: Those that are created fully from ODbL, those
> that are fully from CC-BY-SA, and those that have ODbL data but CC-BY-SA
> coastline. All three would be CC-BY-SA (although the first wouldn't have
> to be), but all three have different attribution requirements.
> 
> I think we should be pragmatic here along the lines of what I said in
> the first paragraph.

Yes, that would work well. A legally paranoid company could still delete old
tiles if they were worried.

> > What I see as the most practical solution is to release the first new
> > planet as dual-licensed and release the next one as ODbL only. This
> > would give data consumers a week to import the new planet, render some
> > new tiles and delete old tiles, then change their attribution.
> 
> This is an option but maybe it confuses people more than need be? So I
> get this new planet and import it into my database and create tiles from
> it that are (at first) CC-BY-SA under dual-licensing but then at some
> future point in time I flip a switch and say "from now on, all tiles are
> CC-BY-SA under ODbL produced works clause so the attribution needs to
> change, even for those tiles that are still the same as yesterday"?

Maybe release the first new planet as dual-licensed but immediately start
the diffs as ODbL only?

The first planet could be recreated as CC BY-SA data by someone else by them
running the redaction bot themselves so in effect it would be available as
CC BY-SA anyways. It would just require a non-trivial amount of work to
process it.

The more I think about it, the more I like offering the first planet also as
cc by-sa. If OSMF doesn't do this, I may dump my pgsnapshot database to do
the equivalent.

> > I realize that this extends the process by a week but when you're
> > dealing with as much data as OSM has there needs to be a transition
> > period for consumers of the data.
> 
> It wouldn't really extend the process by a week because if someone is
> keen on using data under ODbL they could do so immediately.

True - although as I originally envisioned it it would of extended the time
that OSM was available as CC BY-SA, but if only the planet was dual-licensed
then diffs and subsequent planets were not it wouldn't extend the time.