ODbL "virality" questions

[Matt Amos-3]

hi legals,

i've come across a couple of interesting questions / use-cases for the
ODbL and wider discussion. it basically reduces to whether we want the
ODbL to have viral (GPL-like) behaviour, or whether it should be less
viral (LGPL-like). we've discussed this at an LWG meeting and the
general feeling was that the LGPL-like behaviour would be more
desirable, as it would allow wider use of OSM by third parties.
however, it was felt that a wider discussion is necessary.

first case: a site wishes to use OSM data as a basis for
non-geographic data. the example used is a review side, like
beerintheevening.com or tripadvisor.com. they might want to use OSM as
the source of geographic data by linking its reviews to OSM node IDs
(or lat/lons taken from the OSM data). under a GPL-like interpretation
of the ODbL, this would "taint" the database, requiring its release.
considering that the records in the database may contain private
information (IP/email address of the reviewer) this may mean that the
site decides not to use OSM, because releasing the DB would violate
their own privacy policy.

second case: OSM data is downloaded to a handheld device (e.g:
iphone). this is likely (given the screen size of the device) to be an
insubstantial amount. the data is locally used for reference when
entering other information (e.g: abovesaid reviews). the reviews are
uploaded to a non-OSM site, linked to the OSM-derived node ID or
lat/lon. if many people do this, does that constitute repeated
extraction and therefore require release of the non-OSM DB under the
ODbL? i.e: can 3rd party sites use OSM IDs or lat/lons from OSM as
keys into their database?

the discussion at the LWG meeting centered around whether the database
"linking to" OSM data could be considered stand-alone. using the
similarity with the LGPL; whether the reviews database could be
"re-linked" against another source of geographic data while continuing
to work. this would imply that the list of (e.g: pubs or hotels) would
need to be released as an extract of OSM as a list of OSM IDs or
lat/lons, but that the reviews themselves and auxillary tables (such
as the users' information) wouldn't constitute a derivative work of
the OSM database.

what are your thoughts?

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Laurence Penney-2]

Interesting stuff, Matt.

Back in 2005 I asked Ed Parsons, then still at Ordnance Survey, a  
similar question about OS TOIDs. I wondered if I could use TOIDs as  
tags in my own database of photos, in the sense "photo x depicts TOID  
y" - of course omitting all location data associated with the TOID in  
MasterMap. The method of acquiring the TOIDs wasn't specified. He  
didn't know the answer, and to my proposal that a Freedom of  
Information request might be one way for me to get the data out of the  
OS, he suggested I give it a go.

So I did. For my test case I asked for a list of all churches in  
London, along with their TOIDs. I got the response:

“Whilst I can confirm that Ordnance Survey does hold this information,  
I regret to inform you that your request falls within the ‘Formats of  
documents’ exclusion under section 11.(3)(a) of the Re-use of Public  
Sector Information Regulations 2005 (PSI) whereby we are not obliged  
to adapt information to comply with a request. This exclusion applies  
because even though the information exists, it does do as a part of  
the Ordnance Survey MasterMap topography layer and is not available  
through us as separate information. Additionally, under section 21 of  
the Freedom of Information Act 2000 (FOIA), we would not release the  
information as it is ‘accessible to the applicant (you) by other  
means’. This same reason applies under section 5.(2)(c) of PSI.”

It seemed clear that such data extractions would not be considered  
public domain, simply by virtue of having no grid reference or lat-
long. They were part of MasterMap, hence regarded as chargeable data.

Even if the OS had responded with the data, often forgotten is the  
fact that any data provided in an FoI responses is, by default, under  
crown copyright. See this notorious case:

http://timworstall.typepad.com/timworstall/2006/07/craig_murray_an.html

So even if they had responded with the data, I probably wouldn't have  
been able to anything with it. (A local authority might well respond  
positively to an FOIA request of, say, a list of all the footbridges  
in its jurisdiction, yet I'd not necessarily be allowed to republish  
that data, or use the TOIDs in my database.)

At the Society of Cartographers conference last month Peter Miller put  
a closely related question to Vanessa Lawrence:

“Is it possible to, and I have had some indications that it is  
possible, to create a database that basically says ‘this feature in  
Ordnance Survey is the same as this feature in all of these other  
databases [Teleatlas, Navteq, OSM] such that an asset which is  
collected and associated with a feature in one of them can be used in  
features in the other ones as long as the person who provides that  
information is happy for that to be the case.”

After some apparent confusion about whether Peter was asking a  
technical or a legal question [the latter was the case], Vanessa  
responded: “I’m afraid I have no idea. If that’s a question you’d like  
to pose to us, please write in.” (Any response yet, Peter?)

It’s good that OSM is asking the same questions of itself!

FWIW, I very much hope that OSM would be freer with its IDs than  
Ordnance Survey seems to be with its TOIDs. However, since Vanessa had  
“no idea” about the OS's policy on TOID reuse, perhaps there isn’t one.

- L

On 2 Oct 2009, at 18:06, Matt Amos wrote:

> hi legals,
>
> i've come across a couple of interesting questions / use-cases for the
> ODbL and wider discussion. it basically reduces to whether we want the
> ODbL to have viral (GPL-like) behaviour, or whether it should be less
> viral (LGPL-like). we've discussed this at an LWG meeting and the
> general feeling was that the LGPL-like behaviour would be more
> desirable, as it would allow wider use of OSM by third parties.
> however, it was felt that a wider discussion is necessary.
>
> first case: a site wishes to use OSM data as a basis for
> non-geographic data. the example used is a review side, like
> beerintheevening.com or tripadvisor.com. they might want to use OSM as
> the source of geographic data by linking its reviews to OSM node IDs
> (or lat/lons taken from the OSM data). under a GPL-like interpretation
> of the ODbL, this would "taint" the database, requiring its release.
> considering that the records in the database may contain private
> information (IP/email address of the reviewer) this may mean that the
> site decides not to use OSM, because releasing the DB would violate
> their own privacy policy.
>
> second case: OSM data is downloaded to a handheld device (e.g:
> iphone). this is likely (given the screen size of the device) to be an
> insubstantial amount. the data is locally used for reference when
> entering other information (e.g: abovesaid reviews). the reviews are
> uploaded to a non-OSM site, linked to the OSM-derived node ID or
> lat/lon. if many people do this, does that constitute repeated
> extraction and therefore require release of the non-OSM DB under the
> ODbL? i.e: can 3rd party sites use OSM IDs or lat/lons from OSM as
> keys into their database?
>
> the discussion at the LWG meeting centered around whether the database
> "linking to" OSM data could be considered stand-alone. using the
> similarity with the LGPL; whether the reviews database could be
> "re-linked" against another source of geographic data while continuing
> to work. this would imply that the list of (e.g: pubs or hotels) would
> need to be released as an extract of OSM as a list of OSM IDs or
> lat/lons, but that the reviews themselves and auxillary tables (such
> as the users' information) wouldn't constitute a derivative work of
> the OSM database.
>
> what are your thoughts?
>
> cheers,
>
> matt
>
> _______________________________________________
> legal-talk mailing list
> legal-talk@...
> http://lists.openstreetmap.org/listinfo/legal-talk
>

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On 10/5/09, Laurence Penney <lorp@...> wrote:
> It seemed clear that such data extractions would not be considered
> public domain, simply by virtue of having no grid reference or lat-
> long. They were part of MasterMap, hence regarded as chargeable data.

that's the suck-'em-dry licensing model ;-)

> So even if they had responded with the data, I probably wouldn't have
> been able to anything with it. (A local authority might well respond
> positively to an FOIA request of, say, a list of all the footbridges
> in its jurisdiction, yet I'd not necessarily be allowed to republish
> that data, or use the TOIDs in my database.)

FOIA, for some values of "free"...

> It’s good that OSM is asking the same questions of itself!
>
> FWIW, I very much hope that OSM would be freer with its IDs than
> Ordnance Survey seems to be with its TOIDs. However, since Vanessa had
> “no idea” about the OS's policy on TOID reuse, perhaps there isn’t one.

i would hope so too, as it makes OSM data more attractive for those
users who don't need to manipulate the data, but need to annotate it
or reference it. i, for one, would really like to see the next
beerintheevening or tripadvisor based on OSM data, not just the tiles.

we have the opportunity here to decide whether or not we, as a
community, feel that this use of OSM data is OK. from my reading of
the ODbL (insert standard disclaimers here) it's a grey area which we
can strongly influence by public discussion.

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Frederik Ramm]

Hi,

(for those on dev; this started out as a discussion on whether or not we
want to put any legal/license restrictions on external users linking to
OSM objects for identification, e.g. a restaurant guide saying "this pub
is OSM node #12345")

Matt Amos wrote:
> i would hope so too, as it makes OSM data more attractive for those
> users who don't need to manipulate the data, but need to annotate it
> or reference it. i, for one, would really like to see the next
> beerintheevening or tripadvisor based on OSM data, not just the tiles.

My problem with this is that while I'd gladly allow anybody to do that
from a licensing point of view, I'd rather not have people do that from
a technical point of view (and that's why I'm switching over to dev with
this).

Personally I view OSM object IDs as quite frail, and subject to change
without notice at any time. I do not think that OSM object IDs should be
used as foreign keys in any application. I even object to all those
lists on the Wiki which point to hard-coded relation IDs - I, for one,
will delete and re-create an object any time without much thought if it
makes sense to me, breaking any such external entry point.

I fear that if many people treat OSM IDs as permanent, this will have a
restraining influence on us editing our own data ("I wanted to replace
this restaurant node by a building outline for the restaurant but then I
got complaints from users of 15 restaurant guides and Flickr because the
restaurant had suddenly vanished there, and so I reverted my edit").

Until now, if someone asked me a question in that direction, I always
said they should make their own ID a foreign key and tag the OSM object
with something like "restaurant_guide_xyz_id:1234". Which is not ideal
from a data access perspective of course, but allows people editing OSM
to properly work with that.

If we really want to head in a direction where external users refer to
OSM objects, then I think it would be wise to manifest that in the
database somehow, and create some kind of "permanence API" or so, where
you can request a permanent handle for a certain object from the API,
and the API will give you a number, and then if someone deletes and
re-creates an object they will be able to transfer that number to the
new object somehow.

This is of course something for the future, but letting external users
refer directly to OSM IDs sounds like asking for trouble to me.

Bye
Frederik

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Andrew Turner-8]

> On 2 Oct 2009, at 18:06, Matt Amos wrote:
>
>> hi legals,
>>
>> i've come across a couple of interesting questions / use-cases for the
>> ODbL and wider discussion. it basically reduces to whether we want the
>> ODbL to have viral (GPL-like) behaviour, or whether it should be less
>> viral (LGPL-like). we've discussed this at an LWG meeting and the
>> general feeling was that the LGPL-like behaviour would be more
>> desirable, as it would allow wider use of OSM by third parties.
>> however, it was felt that a wider discussion is necessary.
>>
>> first case: a site wishes to use OSM data as a basis for
>> non-geographic data. the example used is a review side, like
>> beerintheevening.com or tripadvisor.com. they might want to use OSM as
>> the source of geographic data by linking its reviews to OSM node IDs
>> (or lat/lons taken from the OSM data). under a GPL-like interpretation
>> of the ODbL, this would "taint" the database, requiring its release.
>> considering that the records in the database may contain private
>> information (IP/email address of the reviewer) this may mean that the
>> site decides not to use OSM, because releasing the DB would violate
>> their own privacy policy.
>>
>> second case: OSM data is downloaded to a handheld device (e.g:
>> iphone). this is likely (given the screen size of the device) to be an
>> insubstantial amount. the data is locally used for reference when
>> entering other information (e.g: abovesaid reviews). the reviews are
>> uploaded to a non-OSM site, linked to the OSM-derived node ID or
>> lat/lon. if many people do this, does that constitute repeated
>> extraction and therefore require release of the non-OSM DB under the
>> ODbL? i.e: can 3rd party sites use OSM IDs or lat/lons from OSM as
>> keys into their database?
>

I've had these very same questions. The in-person responses have
typically been "of course that's ok to do without releasing the review
data" but never in any way that I thought would make a large company
feel comfortable.

I understand that typically copyright law like this is at the behest
of 'best practices' and prior cases - but obviously this is not the
model OSM follows in general and is in fact trying to break out of.

Really, it is akin to linking to a URL (if you consider any node in
OSM is a Resource and could have a URI). My linking to a Wikipedia
definition of "Map" does not change the copyright of my material.

What can we do to make it very clear if this is acceptable use of OSM?
Can we make it very clear that the equivalent of 'linking' to OSM data
that doesn't alter it (or effectively replace primary data within OSM)
does not virally release all data linked to it?

Thanks,
Andrew

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On Tue, Oct 6, 2009 at 1:42 PM, Andrew Turner
<ajturner@...> wrote:

>> On 2 Oct 2009, at 18:06, Matt Amos wrote:
>>
>>> hi legals,
>>>
>>> i've come across a couple of interesting questions / use-cases for the
>>> ODbL and wider discussion. it basically reduces to whether we want the
>>> ODbL to have viral (GPL-like) behaviour, or whether it should be less
>>> viral (LGPL-like). we've discussed this at an LWG meeting and the
>>> general feeling was that the LGPL-like behaviour would be more
>>> desirable, as it would allow wider use of OSM by third parties.
>>> however, it was felt that a wider discussion is necessary.
>>>
>>> first case: a site wishes to use OSM data as a basis for
>>> non-geographic data. the example used is a review side, like
>>> beerintheevening.com or tripadvisor.com. they might want to use OSM as
>>> the source of geographic data by linking its reviews to OSM node IDs
>>> (or lat/lons taken from the OSM data). under a GPL-like interpretation
>>> of the ODbL, this would "taint" the database, requiring its release.
>>> considering that the records in the database may contain private
>>> information (IP/email address of the reviewer) this may mean that the
>>> site decides not to use OSM, because releasing the DB would violate
>>> their own privacy policy.
>>>
>>> second case: OSM data is downloaded to a handheld device (e.g:
>>> iphone). this is likely (given the screen size of the device) to be an
>>> insubstantial amount. the data is locally used for reference when
>>> entering other information (e.g: abovesaid reviews). the reviews are
>>> uploaded to a non-OSM site, linked to the OSM-derived node ID or
>>> lat/lon. if many people do this, does that constitute repeated
>>> extraction and therefore require release of the non-OSM DB under the
>>> ODbL? i.e: can 3rd party sites use OSM IDs or lat/lons from OSM as
>>> keys into their database?
>>
>
> I've had these very same questions. The in-person responses have
> typically been "of course that's ok to do without releasing the review
> data" but never in any way that I thought would make a large company
> feel comfortable.
>
> I understand that typically copyright law like this is at the behest
> of 'best practices' and prior cases - but obviously this is not the
> model OSM follows in general and is in fact trying to break out of.
>
> Really, it is akin to linking to a URL (if you consider any node in
> OSM is a Resource and could have a URI). My linking to a Wikipedia
> definition of "Map" does not change the copyright of my material.
>
> What can we do to make it very clear if this is acceptable use of OSM?
> Can we make it very clear that the equivalent of 'linking' to OSM data
> that doesn't alter it (or effectively replace primary data within OSM)
> does not virally release all data linked to it?

we can make this very clear by:
1) discussing it here,
2) forming a consensus,
3) documenting the results (e.g: on the license FAQ on the wiki)

when i have discussed the question of grey areas with lawyers the
answer has always been that we can make up our own rules in those
areas. the prerequisites for it having a good chance of standing up in
court are that it is easy to find these guidelines in association with
the license and that it represents a consensus view of the community.

so far, all the responses seem to indicate that everyone thinks
linking to OSM data by ID is OK. what about Andy's idea, though? is it
OK to take a location, name and possibly an ID as well to perform
"fuzzy" linking?

my view is that all the linked-to OSM information would have to be
released; the list of (location, name, ID) tuples. but that it would
still be OK to not release the linked-by proprietary information.

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[James Livingston-2]

On 06/10/2009, at 11:30 PM, Matt Amos wrote:
> so far, all the responses seem to indicate that everyone thinks
> linking to OSM data by ID is OK. what about Andy's idea, though? is it
> OK to take a location, name and possibly an ID as well to perform
> "fuzzy" linking?
>
> my view is that all the linked-to OSM information would have to be
> released; the list of (location, name, ID) tuples. but that it would
> still be OK to not release the linked-by proprietary information.

That sounds good in theory, but I think at some point getting out the  
locations and names of things could be "Extraction and Re-utilisation  
of the a Substantial part of the Contents".

Am I allowed to mine the database for the name and location of all the  
pubs and restaurants in the world, without having the data fall under  
the ODbL?  If not, how could it become okay if I claim to just be  
using them as lookup keys? I guess you could have a database with all  
of your proprietary data, and second one which acts as a link between  
the fuzzy-OSM data and IDs in your database, and only release the  
second.

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On Tue, Oct 6, 2009 at 2:55 PM, James Livingston <doctau@...> wrote:

> On 06/10/2009, at 11:30 PM, Matt Amos wrote:
>> so far, all the responses seem to indicate that everyone thinks
>> linking to OSM data by ID is OK. what about Andy's idea, though? is it
>> OK to take a location, name and possibly an ID as well to perform
>> "fuzzy" linking?
>>
>> my view is that all the linked-to OSM information would have to be
>> released; the list of (location, name, ID) tuples. but that it would
>> still be OK to not release the linked-by proprietary information.
>
> That sounds good in theory, but I think at some point getting out the
> locations and names of things could be "Extraction and Re-utilisation
> of the a Substantial part of the Contents".

yes, i think it is.

> Am I allowed to mine the database for the name and location of all the
> pubs and restaurants in the world, without having the data fall under
> the ODbL?  If not, how could it become okay if I claim to just be
> using them as lookup keys? I guess you could have a database with all
> of your proprietary data, and second one which acts as a link between
> the fuzzy-OSM data and IDs in your database, and only release the
> second.

no-one is suggesting that the extraction of names, locations and IDs
would be somehow outside of the ODbL. any site using these as lookup
keys would have to release that data under the ODbL. the question is,
should they have to further release the data in their database which
is using those lookup keys?

as a concrete example, let's pretend i have a site, beerintheOSM,
which rates pubs and allows commenting and photo uploads. if i'm
storing the reviews linked against pubs linked against OSM
(name/location/ID), i definitely have to release the
(name/location/ID) records - that's not up for discussion. should i
also have to release the reviews, comments and photos records despite
the fact that they have no OSM-derived data in them? should i have to
release my entire database, including my users table?

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[IamPersistent]

Matt Amos wrote:
> as a concrete example, let's pretend i have a site, beerintheOSM,
> which rates pubs and allows commenting and photo uploads. if i'm
> storing the reviews linked against pubs linked against OSM
> (name/location/ID), i definitely have to release the
> (name/location/ID) records - that's not up for discussion. should i
> also have to release the reviews, comments and photos records despite
> the fact that they have no OSM-derived data in them? should i have to
> release my entire database, including my users table?
>  
I sincerely hope this is not the case.  It would greatly restrict what
people could do with the data.  That really doesn't sound like "free"
data, if that is the case.

Richard

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Frederik Ramm]

Hi,

> no-one is suggesting that the extraction of names, locations and IDs
> would be somehow outside of the ODbL. any site using these as lookup
> keys would have to release that data under the ODbL.

[...]

> as a concrete example, let's pretend i have a site, beerintheOSM,
> which rates pubs and allows commenting and photo uploads. if i'm
> storing the reviews linked against pubs linked against OSM
> (name/location/ID), i definitely have to release the
> (name/location/ID) records

Wait a minute.

If I run "beerintheOSM" as a crowdourced project - say, a Wiki - and
people can enter new pubs, and the names are entered by those who create
the entries, and I don't even store lat/lon locations, I just allow my
users to add an OSM node id in some kind of template which I then use to
retrieve and display the map for the area, then surely I do not have to
release the records?

* The name was not taken from OSM
* the location is not even stored in my database
* the OSM ID... well yes this would have to be released but not with
context, i.e. I could simply release a list of OSM IDs saying "these are
used in beerintheOSM somewhere

If anyone doubts the above then think what would happen if I didn't use
the OSM ID to draw a map, instead the OSM ID would just be listed there
in the text ("by the way, this pub is OSM node #1234") - which is the
same from a database perspective. Surely such reference cannot trigger
any viral effect?

Bye
Frederik


_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Dan Karran]

2009/10/6 Matt Amos <zerebubuth@...>:

> as a concrete example, let's pretend i have a site, beerintheOSM,
> which rates pubs and allows commenting and photo uploads. if i'm
> storing the reviews linked against pubs linked against OSM
> (name/location/ID), i definitely have to release the
> (name/location/ID) records - that's not up for discussion. should i
> also have to release the reviews, comments and photos records despite
> the fact that they have no OSM-derived data in them? should i have to
> release my entire database, including my users table?

I would say that related information like reviews, comments, etc. that
were added to beerintheOSM shouldn't need to be released, but we
should encourage any information that *could* live in OSM (e.g. if
users added opening hours) to be released so it could be
re-incorporated.

What would happen if the beerintheOSM site encouraged their users to
add new pubs to their site, would that data - the equivalent of what
would have come from OSM, had they come from there - need to be
released as well, or again something we should just encourage the site
to release?


Dan

--
Dan Karran
dan@...
www.dankarran.com

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On Wed, Oct 7, 2009 at 12:14 AM, Frederik Ramm <frederik@...> wrote:

> Hi,
>
>> as a concrete example, let's pretend i have a site, beerintheOSM,
>> which rates pubs and allows commenting and photo uploads. if i'm
>> storing the reviews linked against pubs linked against OSM
>> (name/location/ID), i definitely have to release the
>> (name/location/ID) records
>
> Wait a minute.
>
> If I run "beerintheOSM" as a crowdourced project - say, a Wiki - and
> people can enter new pubs, and the names are entered by those who create
> the entries, and I don't even store lat/lon locations, I just allow my
> users to add an OSM node id in some kind of template which I then use to
> retrieve and display the map for the area, then surely I do not have to
> release the records?
>
> * The name was not taken from OSM
> * the location is not even stored in my database
> * the OSM ID... well yes this would have to be released but not with
> context, i.e. I could simply release a list of OSM IDs saying "these are
> used in beerintheOSM somewhere

this is the crux of the question. the ODbL makes no distinction
between lat/lon data, ID data, or any other sort of data. so the
question then becomes; if i'm using some data from an ODbL database
and incorporating that into my database, do i have to release all of
my database, or just the bits of it which came from or were derived
from the ODbL database?

> If anyone doubts the above then think what would happen if I didn't use
> the OSM ID to draw a map, instead the OSM ID would just be listed there
> in the text ("by the way, this pub is OSM node #1234") - which is the
> same from a database perspective. Surely such reference cannot trigger
> any viral effect?

that reference isn't any different from any other datum from the
database, hence the question ;-)

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On Wed, Oct 7, 2009 at 12:21 AM, Dan Karran <dan@...> wrote:
> What would happen if the beerintheOSM site encouraged their users to
> add new pubs to their site, would that data - the equivalent of what
> would have come from OSM, had they come from there - need to be
> released as well, or again something we should just encourage the site
> to release?

good question. even if OSM IDs or lat/lons are OK for linking
purposes, where do we draw the line on what is "attached" data, like
reviews, and what is new data added to the ODbL extract, requiring it
to be available?

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Frederik Ramm]

Hi,

Matt Amos wrote:
> this is the crux of the question. the ODbL makes no distinction
> between lat/lon data, ID data, or any other sort of data. so the
> question then becomes; if i'm using some data from an ODbL database
> and incorporating that into my database, do i have to release all of
> my database, or just the bits of it which came from or were derived
> from the ODbL database?

Let's look at the reason why we have this whole viral license, shall we?

(I'm taking off my "this is all stupid and we should do PD" hat for a
moment and act as if I were a share-aliker.)

The idea behind this is that we don't want to give anything to people
which they then make proprietary - the worst case being that one day OSM
ceases to exist and only some proprietary copy remains. The license is
there to ensure that OSM data remains free.

But a site that *only* takes OSM IDs in order to link to them does not
create anything of their own. If OSM one day ceases to exist then the
OSM IDs stored in that site become worthless. They only store pointers
into our database, they don't make a derived product. (If I tell you to
download a film and skip to 6'32 because that's where the action is, am
I creating a "derived work" of that film?)

You are right in saying that by the letter of the license, an id is data
just like anything else. But the spirit surely affects only *useful* data?

Bye
Frederik

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On Wed, Oct 7, 2009 at 3:59 PM, Frederik Ramm <frederik@...> wrote:

> Hi,
>
> Matt Amos wrote:
>> this is the crux of the question. the ODbL makes no distinction
>> between lat/lon data, ID data, or any other sort of data. so the
>> question then becomes; if i'm using some data from an ODbL database
>> and incorporating that into my database, do i have to release all of
>> my database, or just the bits of it which came from or were derived
>> from the ODbL database?
>
> Let's look at the reason why we have this whole viral license, shall we?
>
> (I'm taking off my "this is all stupid and we should do PD" hat for a
> moment and act as if I were a share-aliker.)

/me adjusts headgear also.

> The idea behind this is that we don't want to give anything to people
> which they then make proprietary - the worst case being that one day OSM
> ceases to exist and only some proprietary copy remains. The license is
> there to ensure that OSM data remains free.
>
> But a site that *only* takes OSM IDs in order to link to them does not
> create anything of their own. If OSM one day ceases to exist then the
> OSM IDs stored in that site become worthless. They only store pointers
> into our database, they don't make a derived product. (If I tell you to
> download a film and skip to 6'32 because that's where the action is, am
> I creating a "derived work" of that film?)

no, but a time code pointer into a copyrighted work isn't necessarily
a good analogy for an ID extracted from a database rights-protected
database.

> You are right in saying that by the letter of the license, an id is data
> just like anything else. But the spirit surely affects only *useful* data?

i think i can hear lawyers having heart attacks over the word "useful" ;-)

can the SA requirement be satisfied by saying that we consider the
extracted IDs to be an ODbL part of a collective database, where the
proprietary data is the other part? it would require the ODbL part
(i.e: the list of IDs) to be made available, but nothing else.

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Frederik Ramm]

Hi,

Matt Amos wrote:
> can the SA requirement be satisfied by saying that we consider the
> extracted IDs to be an ODbL part of a collective database, where the
> proprietary data is the other part? it would require the ODbL part
> (i.e: the list of IDs) to be made available, but nothing else.

It would work, but I'm trying to think if this would have adverse side
effects.

Can this be compared to Google importing all of OSM into MapMaker and
then only making available the OSM part and not anything newly created?
Would that still count as a collective database? If not, then where is
the boundary?

I'm going back to this notion of "usefulness". Firstly, is a database a
database if it is one-dimensional? My feeling is that a database must
always combine at least two values: Have one, look up the other; have
the other, look up the first. Is a list of all valid post codes in a
country still a database if it doesn't have names or geometries? Is the
list of all latitudes in OSM still a database?

It could be that an extract of some OSM IDs is not even a derived database.

Assuming for a moment it were a database, then, being rather useless, is
it substantial? Could we perhaps say that if you extract only one
dimension from OSM, this can never be a substantial extract - a list of
latitudes, a list of longitudes, a list of keys or a list of values, or
a list of IDs?

Or could we perhaps even specify that anything that doesn't use our
geometry is not substantial? A list of all pubs in Madrid would be
substantial since it needs geometry; a list of all pubs on the planet
would not be substantial. That would neatly cover anyone wanting to use
any number of OSM IDs for linking as he would never use the geometry
from OSM.

But that would again raise a cascading substantial-ness problem - what
if I publish an OSM extract of Madrid and someone else counts all pubs
in there.

Bye
Frederik

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On 10/7/09, Frederik Ramm <frederik@...> wrote:

> Hi,
>
> Matt Amos wrote:
>> can the SA requirement be satisfied by saying that we consider the
>> extracted IDs to be an ODbL part of a collective database, where the
>> proprietary data is the other part? it would require the ODbL part
>> (i.e: the list of IDs) to be made available, but nothing else.
>
> It would work, but I'm trying to think if this would have adverse side
> effects.
>
> Can this be compared to Google importing all of OSM into MapMaker and
> then only making available the OSM part and not anything newly created?
> Would that still count as a collective database? If not, then where is
> the boundary?

yep, that's the question ;-)

> I'm going back to this notion of "usefulness". Firstly, is a database a
> database if it is one-dimensional? My feeling is that a database must
> always combine at least two values: Have one, look up the other; have
> the other, look up the first. Is a list of all valid post codes in a
> country still a database if it doesn't have names or geometries? Is the
> list of all latitudes in OSM still a database?

possibly. if we consider the IDs (or any other one dimension) to be
insubstantial, then this might work. the problem comes when using the
scheme that Andy suggested. it may not be practical to link to a
single dimension of an OSM dataset - it might be necessary to extract
multiple dimensions to do good fuzzy matching.

> It could be that an extract of some OSM IDs is not even a derived database
>
> Assuming for a moment it were a database, then, being rather useless, is
> it substantial? Could we perhaps say that if you extract only one
> dimension from OSM, this can never be a substantial extract - a list of
> latitudes, a list of longitudes, a list of keys or a list of values, or
> a list of IDs?

from the ODbL:

“Derivative Database” – Means a database based upon the Database, and
includes any translation, adaptation, arrangement, modification, or
any other alteration of the Database or of a Substantial part of the
Contents. This includes, but is not limited to, Extracting or
Re-utilising the whole or a Substantial part of the Contents in a new
Database.

so, if we consider ID extraction to be substantial, then the answer is
definitely yes - it would be an arrangement of a substantial part of
the contents. if we consider IDs to be insubstantial then it would be
OK.

special note: the word "new" in the above definition means "other or
different from the first", not "created from scratch".

> Or could we perhaps even specify that anything that doesn't use our
> geometry is not substantial? A list of all pubs in Madrid would be
> substantial since it needs geometry; a list of all pubs on the planet
> would not be substantial. That would neatly cover anyone wanting to use
> any number of OSM IDs for linking as he would never use the geometry
> from OSM.

are you suggesting that someone wanting to run beerintheOSM would have
to have a worldwide scope? it wouldn't even be possible to be
country-specific because that would give it a geographic scope and
therefore depend on the geometry?

> But that would again raise a cascading substantial-ness problem - what
> if I publish an OSM extract of Madrid and someone else counts all pubs
> in there.

i think counts, or any other summary statistic of the non-identifiable
sort (i.e: that which would pass privacy regulations), could be
considered insubstantial. the number of pubs in any given large-enough
bounding box, certainly a city or country, shouldn't be considered a
substantial extract.

in any case, the cascading substantialness is already taken care of:
if my dataset A is substantial and you derive B from that, the same
standards of substantialness apply to your extract of B from A as they
do to my extract of A from OSM. if A is insubstantial, then it would
be impossible to make a substantial extract from that insubstantial
extract. or so the theory goes ;-)

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Frederik Ramm]

Hi,

Matt Amos wrote:

>> Or could we perhaps even specify that anything that doesn't use our
>> geometry is not substantial? A list of all pubs in Madrid would be
>> substantial since it needs geometry; a list of all pubs on the planet
>> would not be substantial. That would neatly cover anyone wanting to use
>> any number of OSM IDs for linking as he would never use the geometry
>> from OSM.
>
> are you suggesting that someone wanting to run beerintheOSM would have
> to have a worldwide scope? it wouldn't even be possible to be
> country-specific because that would give it a geographic scope and
> therefore depend on the geometry?

I was thinking that if he relies on *our* geometry then he's making a
substantial extract, whereas if he uses some other means to list the
pubs in, say, England and then just references our nodes, that's ok then.

In one case he has the data already (name of pub + as much knowledge
about the location as required for his application) and only links to
OSM as an additional source of info. In the other case he is using OSM
to find information in the first place.

Bye
Frederik

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Matt Amos-3]

On 10/8/09, Frederik Ramm <frederik@...> wrote:

> Hi,
>
> Matt Amos wrote:
>>> Or could we perhaps even specify that anything that doesn't use our
>>> geometry is not substantial? A list of all pubs in Madrid would be
>>> substantial since it needs geometry; a list of all pubs on the planet
>>> would not be substantial. That would neatly cover anyone wanting to use
>>> any number of OSM IDs for linking as he would never use the geometry
>>> from OSM.
>>
>> are you suggesting that someone wanting to run beerintheOSM would have
>> to have a worldwide scope? it wouldn't even be possible to be
>> country-specific because that would give it a geographic scope and
>> therefore depend on the geometry?
>
> I was thinking that if he relies on *our* geometry then he's making a
> substantial extract, whereas if he uses some other means to list the
> pubs in, say, England and then just references our nodes, that's ok then.

does that mean it's not ok to look at the lat/lon to find the list of
pubs, but it's ok to look at the name tag? is that still true if i
produce a planet derivative which automatically adds is_in tags based
on the administrative boundary data - he can use that is_in tag?

> In one case he has the data already (name of pub + as much knowledge
> about the location as required for his application) and only links to
> OSM as an additional source of info. In the other case he is using OSM
> to find information in the first place.

i think the more useful case to most people will be to use the OSM
data geographically. if i started beerintheOSM i'd want to use OSM for
as much of the geographic data as possible - that's kinda the point of
OSM isn't it?

so, assuming beerintheOSM has a list of IDs, names and locations of
pubs, let's say extracted by xapi query using the UK bbox, does that
mean it would have to release its whole database, or just the
OSM-derived parts of that database?

cheers,

matt

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk

[Erik Johansson-10]

On Thu, Oct 8, 2009 at 10:33 AM, Matt Amos <zerebubuth@...> wrote:
> i think the more useful case to most people will be to use the OSM
> data geographically. if i started beerintheOSM i'd want to use OSM for
> as much of the geographic data as possible - that's kinda the point of
> OSM isn't it?
>
> so, assuming beerintheOSM has a list of IDs, names and locations of

Just to be clear you assume that there are no geodata that isn't in
the OSM database? If there is also a non OSM database with lat/lon for
pubs then that should be share-alike, but for the stuff that isn't
geodata I can't really say

I don't agree with Frederik. Using OSM ID will link your database in
the same way as linking shared libraries does. Hence making your
database share-alike, again the question is what data should be
share-alike.


/erik

_______________________________________________
legal-talk mailing list
legal-talk@...
http://lists.openstreetmap.org/listinfo/legal-talk