[OSM-talk] Critical Mass for license change-over
frederik at remote.org
Sat Jan 28 12:10:58 GMT 2012
On 01/28/2012 05:59 AM, Michal Migurski wrote:
>> I think that realistically, taking into account the time, manpower,
>> and other resources available, you can expect to have an
>> unambiguous plan in the form of a verbal description, or *maybe* at
>> most a script or program that enables you to generate an ODbL
>> planet from the full history file*. But certainly not a definitive,
>> fast, and planet-wide "cleanmap", nor regular planet dumps with the
>> license change rules applied.
> That's weird. What's going to happen on April 1, then?
It is not yet clear what form the algorithm will take. Technically, it
could be something that takes a full history planet file or some other,
specially created, extract and loads it into a new, initially empty,
database; or it could be something that looks at (a copy of) the current
database and selectively deletes stuff from that.
The code used by OSMF to make the switch can certainly be published but
it is not necessarily something that you can quickly run on your
machine; it is possible that the code is geared towards the OSM database
in a way that makes it impractical for someone else to execute it.
Your message seems to imply that someone who is ready to make the
modifications necessary for the license change to our database would
automatically also have to have the technology to create something like
a planet-wide "cleanmap" or planet dumps with the license change rules
applied. This assumption is not true; even if the algorithm is ready, a
planet-wide cleanmap would possibly require a full copy of the current
database to be made on a separate server, dumps to be created from that
and imported into a third server where they can be rendered.
> Keeping in
> mind that I am in support of the license switch, I think it's
> completely reasonable to expect a technical plan for a switch just 60
> days in the future.
You talk about reasonable - I talk about realistic.
> Especially in the context of a thread starting by
> the license group looking for feedback. The question was "do we have
> critical mass?" - there's no way to answer that without a way to
> measure impact.
It is worth noting that the "1st April" goal was set by the OSMF board
in their latest face-to-face meeting, not by the License Working Group.
I am not party to these communications but I believe that the first LWG
heard of that date was after the board meeting was over. I doubt that
LWG have even been consulted beforehand. The first LWG meeting after
that has the following in its minutes:
"Board would like to set 1st April 2012 for cutting over to ODbL latest.
LWG feels that with current status this is a practical goal for the
community to work towards and resolves to meet this target..."
The f2f meeting at which the "1st April" goal was set seems to have been
operating under the headline: "Strategic planning for the coming year.
Set high levels goals, align with some specific 'Big Audacious' actions
for the Foundation...".
>> I agree these things would be nice to have but I don't see where
>> they should come from. Currently we don't even have the algorithm.
> Then it sounds like nobody's ready for April; not the LWG, not the
> Foundation, and not any of us.
Maybe that's why they called it "Big Audacious" ;). Sadly the OSMF board
meeting minutes don't record who came up with 1st April and who
supported the idea, else we could invite these individuals to discuss.
>> If anyone has the hardware and time and brain capacity to build
>> something that generates "parallel planet files", my recommendation
>> is to start setting this up now, even though the final algorithm
>> might not be clear, so that once the algorithm is published you can
>> react quickly.
> I donated money towards a new server just a short time ago. Might
> that be useful for this purpose?
That's for OWG to decide but I don't think the new server is available
> there should be a
> parallel data and tile service set prior to launch and an old data
> set and tile service post-launch.
I think this depends on how the changeover is done. I certainly don't
see an old tile service on the cards but copying over the currently
existing tiles to some static storage should be possible. And an old
data service *might* happen as a side effect *if* the license change
should be done in a fashion where data is loaded onto a new server and
that goes into production - then the old server could carry on read-only
for a while. But it might just as well be that the old database is
dropped and re-created or something, and in that case I don't see anyone
making resources avaialble to carry on serving old data. Of course, if
someone were interested enough, they could just take a planet file and
load it up into a rails port of their own to serve old data.
There's a special mailing list called "rebuild" which has been created
to discuss exactly how the database rebuild is going to be run. Anything
we come up with on that list would have to be tried out of course and
thoroughly examined. I don't think it would be realistic to expect
anything remotely "final" to emerge soon; maybe in a month if you're lucky.
Personally, I don't believe in audacious goals; but then again I believe
that one should at least do what is possible to help. If we don't make
April 1st then we'll want to make May 1st or June 1st, and every minute
we spend making catalogues of things that "someone should do" and things
that "can reasonably be expected" is a minute not spent to actually
achieve these things.
Frederik Ramm ## eMail frederik at remote.org ## N49°00'09" E008°23'33"
More information about the talk