Re: [trdf-prv-vocab-users] Fw: Provenance Vocabulary for Open Street Map

[Olaf H. <ha...@in...>]

Hey Carsten,

Instead of understanding (OSM) change sets as a special kind of creation 
guidelines, it is also possible to understand change sets as a special kind of 
data creations. While an osp:Edit creates a new state of a single feature in 
the OSM dataset, we can understand an osp:Changeset as a prv:DataCreation that 
creates a new state of the whole OSM dataset. Based on this understanding, 
osp:Changeset describes provenance on a different level of granularity as 
osp:Edit does.

Alternatively, I would say that it is not really necessary to map each of the 
terms you introduce to a term in the Provenance Vocabulary. Hence, providing 
no mapping at all for osp:Changeset is also an option ;-)

Greetings,
Olaf


On Monday 14 March 2011 11:40:05 Carsten Keßler wrote:
> The message still waits for approval to get through to the mailing list, so
> I am sending this again directly to you guys.
> 
> Cheers,
> Carsten
> 
> Forwarded message:
> > From: Carsten Keßler <car...@un...>
> > To: trd...@li...
> > Date: Montag, 14. März 2011 11:37:40
> > Subject: Re: [trdf-prv-vocab-users] Provenance Vocabulary for Open Street
> > Map
> > 
> > Hey Jun, hey Olaf,
> > 
> > after reading your discussion, I am also not so sure anymore about
> > modeling the Changeset class as subclass of prv:CreationGuideline.
> > 
> > Olaf's explanations about the relationships between FeatureState, Edit,
> > and Changeset were absolutely correct. If we make Changeset a subclass
> > of prv:CreationGuideline, the question is also, what we gain from this
> > stretching of the original idea behind a creation guideline? Does this
> > buy us anything?
> > 
> > Then again, I am not sure how to align the Changeset class (which we
> > should definitely have, in order to have a complete representation of
> > the OSM structure) with the prv ontology. Any ideas where it might fit
> > better?
> > 
> > Cheers,
> > Carsten
> > 
> > On Freitag, 11. März 2011 at 09:03, Olaf Hartig wrote:
> > > Hey Jun,
> > > 
> > > On Thursday 10 March 2011 14:15:54 Jun Zhao wrote:
> > > > Hi Olaf and Carsten,
> > > > 
> > > > > That's basically all you need. However, you may also want to model
> > > > > your changesets to keep track of which edits belong together. We
> > > > > can understand a changeset as a special kind of
> > > > > prv:CreationGuideline which describes how to perform a
> > > > > prv:DataCreation with which a new state of the complete OSM
> > > > > dataset has been created (even if we do not describe this data
> > > > > creation explicitly). The only thing that's needed is an
> > > > > association of each osp:Changeset with the corresponding osp:Edit
> > > > 
> > > > > instances. Hence, you need a new property:
> > > > I don't understand why a changeset is a creation guileline.
> > > > 
> > > > My understanding of Carsten's scenario is that each feature (state)
> > > > can be associated with a changeset, which is a collection of edit
> > > > events, so that one can figure out what change processes have led to
> > > > the existence of this feature (state).
> > > 
> > > Not exactly. First, a changeset (in the context of OSM) is indeed a
> > > collection of edit events. But -according to my understanding- it is
> > > these edit events that are associated with the states of features (not
> > > the changeset). These edit events can be understood as data creations
> > > which either created a new, first version/state of a feature or which
> > > transformed a specific feature from one version to the next (mostly by
> > > adding/removing properties). The changesets collect such edit events
> > > (for different ! features) that happened at the same time and that
> > > were performed by the same user.
> > > 
> > > > Such a changeset seems like a collection of prv:Executions. And I am
> > > > not sure we have any concepts from the Provenance Vocabulary that
> > > > can exactly capture this concept.
> > > 
> > > You are right with the first sentence - even more, it is a collection
> > > of prv:DataCreations - but -as mentioned before- each of the data
> > > creation in such a collection focuses on a different feature (state).
> > > 
> > > Regarding the question whether it would be correct to understand such a
> > > collection as a prv:CreationGuideline I am a bit undecided (after
> > > reading you concerns). The reason why such an understanding might be
> > > problematic is, that a changeset is a collection of executions (of
> > > processes) that happened; whereas, in a "pure" creation guideline I
> > > would expect descriptions of processes that can be executed. Thus,
> > > understanding a changeset as a creation guideline seems -at least- to
> > > be stretching the limits. However, may be that's still okay from a
> > > pragmatic POV?
> > > 
> > > > And if we take Olaf's suggestion to call each feature at a particular
> > > > state as a featurestate (which is absolutely the right thing to do,
> > > > IMO:), then each such featurestate can be associated with one and
> > > > only one edit execution. I don't see how a featurestate can be
> > > > associated with a changeset anymore, which is a collection of edits.
> > > > If we enforce this strict distinction between features, how do we
> > > > apply changeset to features? Probably I missed something?
> > > 
> > > I guess I already addressed the concerns you raise in that paragraph.
> > > However, you're perfectly right, each feature state can only be
> > > associated with a single edit event - at least in the role of its data
> > > creation (that's why we have the cardinality restriction on
> > > prv:createdBy ;) Then each feature state might also be associated with
> > > a second edit event by which the feature state was used to create the
> > > next state of the corresponding feature (hence, the relationship here
> > > is represented by a prv:usedBy property).
> > > 
> > > Cheers,
> > > Olaf
> > > 
> > > > cheers,
> > > > 
> > > > Jun
> > > > 
> > > > > osp:includesEdit a owl:ObjectProperty
> > > > > 
> > > > >  rdfs:domain osp:Changeset ;
> > > > >  rdfs:range osp:Edit .
> > > > > 
> > > > > Now, you can describe things like:
> > > > > 
> > > > > <http://osm.org/api/0.6/changeset/3610855> a osp:Changeset ;
> > > > > 
> > > > >  osp:includesEdit :edit1 .
> > > > > 
> > > > > Hope that helps,
> > > > > Olaf
> > > > > 
> > > > > On Monday 07 March 2011 10:47:25 Carsten Keßler wrote:
> > > > > > Hi Olaf,
> > > > > > 
> > > > > > thanks for taking the time to look at our extension. See my
> > > > > > answers for your questions below (I'm also CCing this to my
> > > > > > coauthors): 1.) You define a class osp:Feature as a sub-class of
> > > > > > prv:File and from the
> > > > > > 
> > > > > > > other class that you introduce, I see that a feature can be an
> > > > > > > "OSM node with lat/lon coordinates".
> > > > > > 
> > > > > > Either that, or a way or polygon.
> > > > > > 
> > > > > > > I'm not familiar with the OSM terminology. However,
> > > > > > > I doubt that it is correct to understand a point in a
> > > > > > > coordinate system as a file.
> > > > > > 
> > > > > > Well, it is a bit more than a point in a coordinate system; you
> > > > > > can see a more readable serialization of the whole contents by
> > > > > > looking at http://www.openstreetmap.org/browse/node/612986688/
> > > > > > as an example
> > > > > > 
> > > > > > > So, what exactly does a URI like
> > > > > > > http://osm.org/api/0.6/node/612986688/1 identify? The actual
> > > > > > > node? A version of the node? The Web resource with data about
> > > > > > > the node? The Web resource with a specific version of data
> > > > > > > about that node?
> > > > > > 
> > > > > > I think the last one fits best: it is a web resource with a
> > > > > > specific version of data about that node. What you model as a
> > > > > > dataitem (the actual node in our case) is not dereferenceable,
> > > > > > if I understand it correctly. At least not from the outside,
> > > > > > i.e., if you are not running your own OSM server. If you are
> > > > > > running the service, you could link it to an entry in your
> > > > > > database (?)
> > > > > > 
> > > > > > > 2.) Will an HTTP client always get the same representation when
> > > > > > > dereferencing URIs like http://osm.org/api/0.6/node/612986688/1
> > > > > > > ? In other words, does any change create a new version and,
> > > > > > > thus, result in another URI, such as
> > > > > > > http://osm.org/api/0.6/node/612986688/2 ?
> > > > > > 
> > > > > > Yes, exactly.
> > > > > > 
> > > > > > > 3.) You define a class osp:Changeset as a sub-class of
> > > > > > > prv:Artifact; in your example provenance description you have
> > > > > > > instances of osp:Changeset as objects in prv:createdBy
> > > > > > > triples. This makes these instances a prv:DataCreation as
> > > > > > > well. Is that what you intended?
> > > > > > 
> > > > > > Thanks for pointing this out, that's not intended. I'll fix that.
> > > > > > 
> > > > > > > 4.) A few more questions to understand the relationship between
> > > > > > > features (osp:Feature), edits (osp:Edit) and changesets
> > > > > > > (osp:Changeset): * Is the relationship between edits and
> > > > > > > features 1:1, 1:n, or n:1? I.e. does an edit always only
> > > > > > > change a single feature? Can a feature be changed by multiple
> > > > > > > edits?
> > > > > > 
> > > > > > The Edit class has no corresponding element in OSM. We have
> > > > > > introduced it to be able to link a specific change (that was
> > > > > > committed with a Changeset) to the affected feature. Every edit
> > > > > > can only change one feature, the relationship between edits and
> > > > > > features is therefore 1:1.
> > > > > > 
> > > > > > > * Is a changeset some kind of container of edits that have been
> > > > > > > performed at the same time by the same user?
> > > > > > 
> > > > > > Yes, exactly. Most of the OSM editors are based on sessions: you
> > > > > > open the data for a certain map extent in the editor, make you
> > > > > > edits on different features, and finally save everything. This
> > > > > > collection is then uploaded to OSM as a Changeset .
> > > > > > 
> > > > > > Best,
> > > > > > Carsten
> > > > > > 
> > > > > > > Greetings,
> > > > > > > Olaf
> > > > > > > 
> > > > > > > [1] http://dl.dropbox.com/u/9209387/osp-rdf.zip
> > > > 
> > > > ---------------------------------------------------------------------
> > > > ------ --- Colocation vs. Managed Hosting
> > > > A question and answer guide to determining the best fit
> > > > for your organization - today and in the future.
> > > > http://p.sf.net/sfu/internap-sfd2d
> > > > _______________________________________________
> > > > trdf-prv-vocab-users mailing list
> > > > trd...@li...
> > > > https://lists.sourceforge.net/lists/listinfo/trdf-prv-vocab-users