Re: Issue 27: correction of the description of the Coverage in Linked Data deliverable

George, in the context of Linked Data, 'datacube' is often used to refer to
RDF Data Cube [1]; the short name for the supporting vocabulary is QB (as
used by Simon Cox earlier in this thread).

Jeremy

[1]: http://www.w3.org/TR/vocab-data-cube/

On Tue, 13 Oct 2015 at 16:34 Peter Baumann <p.baumann@jacobs-university.de>
wrote:

> let's give it a try and see how much we agree:
>
> - a datacube has an ordered list of d>0 axes
> - a datacube contains cells which are identified (addressed) through a
> vector whose elements are coordinates along each axis, in proper sequence
> - the set of admissible cell positions is the cross product of closed
> intervals along each axis
> - cells hold data items which all belong to one common type ("cell type")
>
> operations:
> - trim and slice
> - Linear Algebra, signal/image processing, statistics
> - geo operations
> - ...and practically motivated: en/decoding using some suitable format
>
> cheers,
> Peter
>
>
>
>
> On 2015-10-13 14:29, George Percivall wrote:
>
> Many organizations use the term “datacube”.  I am not aware of a consensus
> definition.
>
> George
>
>
> On Oct 13, 2015, at 4:09 AM, Peter Baumann <p.baumann@jacobs-university.de>
> wrote:
>
> On 2015-10-13 08:54, Jon Blower wrote:
>
> well, a coverage is a datacube whose axes can be spatial and/or temporal.
>
>
> This is only true for certain types of coverages. Many others (curvilinear
> grids, irregular meshes, polygon-based coverages) don’t fit this definition.
>
>
> you are right, Jon, in that coverages are wider area. Just irregular grids
> are still grids, hence in datacube world.
> I thought I focus for simplicity and blank out what's not in scope here,
> but you caught me ;-)
>
> -Peter
>
>
> But I agree with your wider point that we need to step back and consider
> what our requirements are. I’ve examined QB in a previous project and am
> dubious that it has much practical utility for this kind of thing, but
> that’s only my view from a certain standpoint. We need to define what
> exactly we want to be able to do.
>
> Cheers,
> Jon
>
> On 12 Oct 2015, at 21:41, Peter Baumann <p.baumann@jacobs-university.de>
> wrote:
>
> well, a coverage is a datacube whose axes can be spatial and/or temporal.
> It might be interesting to relate RDF cubes and coverages.
> But again, what do we want to incorporate actually?
> -Peter
>
>
> On 2015-10-12 01:36, Simon.Cox@csiro.au wrote:
>
> I would think that QB[1] (which is derived from SDMX) would have something
> to contribute here. It is an RDF vocabulary that describes the structure of
> a datacube, and this provides specific RDF-oriented queries into cells,
> slices, dimensions of gridded data. Geospatial coverages have the
> additional feature that one or more of the dimensions is spatio-temporal.
>
>
> My view is that there should be no expectation that whole datasets would
> have to be transformed and stored following QB, but that subsets can be
> uniquely identified using QB-bases queries, which would then be transformed
> into the native query (WCS, SOS, OPeNDAP) and passed on to the hosting
> service).
>
>
> [1] http://www.w3.org/TR/vocab-data-cube/
>
>
> *From:* Peter Baumann [mailto:p.baumann@jacobs-university.de
> <p.baumann@jacobs-university.de>]
> *Sent:* Monday, 12 October 2015 8:29 AM
> *To:* Jon Blower <j.d.blower@reading.ac.uk> <j.d.blower@reading.ac.uk>
> *Cc:* Ed Parsons <eparsons@google.com> <eparsons@google.com>; Cox, Simon
> (L&W, Clayton) <Simon.Cox@csiro.au> <Simon.Cox@csiro.au>;
> frans.knibbe@geodan.nl; public-sdw-wg@w3.org
> *Subject:* Re: Issue 27: correction of the description of the Coverage in
> Linked Data deliverable
>
>
> Hi Jon,
>
> exciting questions indeed, you are absolutely right: large portions of the
> overall issue are independent from "to coverage or not to coverage" (sorry
> for bending language).
> What I find particularly interesting is this transition from general data
> linking into referencing the internals of an object. A coverage is just one
> particular case, so solving this might open up vistas for other links -
> into graphs, into documents (I mean: more than just HTML anchors), etc.
> This is one reason why I am curiously following progress in this group.
>
> Nite,
> Peter
>
> On 2015-10-11 21:06, Jon Blower wrote:
>
> Hi Peter,
>
> I’m not suggesting redefining “coverage”, I’m suggesting that there are
> interesting questions around the use of coverages in the Linked Data world
> that aren’t concerned with ISO19123, for example:
>
> 1. Identifying coverages (hence being able to link to them).
> 2. Behaviour of web services that serve coverages (e.g. how can we improve
> WCS, OPeNDAP, NcSS etc to play more nicely with the wider web?).
> 3. Linking between data catalogues and coverage services (e.g. linking
> between GeoDCAT descriptions and concrete data access services)
>
> None of these are within scope for ISO19123, but I believe are interesting
> problems that this group could help with (and are on my mind at the moment
> because we need solutions for the MELODIES project).
>
> The question of linking *into* coverages (i.e. identifying coverage
> subsets) probably does involve stuff like ISO19123(-2), because for that we
> do need some common understanding of what a coverage data structure looks
> like.
>
> By leaping immediately into the ISO19123 world we restrict ourselves
> unnecessarily to the problem of modelling and encoding coverages, which is
> certainly relevant but not the only problem that’s pertinent to Linked Data
> (particularly since there are many other groups covering* some of this).
>
> Cheers,
> Jon
>
> * no pun intended
>
>
>
> On 11 Oct 2015, at 19:29, Peter Baumann <p.baumann@jacobs-university.de>
> wrote:
>
>
> Hi Jon-
>
> several case studies for a range of different areas have been conducted,
> here a theoretical [1] and an applied one [2] - these are just a few, of
> course, others have worked on this, too. It is just that the term
> "coverage" has a particular definition, so we cannot redefine at will if
> interoperability is among the goals. A clear scientific treatment of terms
> seems important. Hence, for scientific groundwork I'd suggest to use a
> neutral term, maybe "pictures" or anything else that appears meaningful and
> not yet taken.
>
> cheers,
> Peter
>
> [1]  Angelica Garcia, Peter Baumann: *Modeling Fundamental Geo-Raster
> Operations with Array Algebra*. IEEE international workshop in spatial
> and spatio-temporal data mining, October 28-31 2007, Omaha, USA
> [2] Peter Baumann, Maximilian Höfner, Walter Schatz: *Querying Large Geo
> Image Databases: A Case Study*. IV Brazilian Symposium on GeoInformatics
> - GeoInfo 2002, December 5-6 2002, Caxambu, Brazil
>
> (BTW, similar studies have been done for astro and life sciences, too)
> On 2015-10-10 20:31, Jon Blower wrote:
>
> Hi all,
>
> I’m relatively new to this group so I don’t know all the history behind
> the wording of the Charter but I have always found this particular
> requirement to be prematurely specific. Personally I would be more
> comfortable with a requirement along the lines of (in imprecise language),
> “We know that a lot of coverage data are being published and such data pose
> challenges for Linked Data approaches. This group will develop
> recommendations for making best use of coverage data in a Linked Data
> environment.”
>
> From this high-level requirement we need to develop specific use cases
> that identify real gaps in the ecosystem and work out what we can actually
> do to fill them, within the scope of this group (and what we defer to other
> groups). I don’t think I’ve seen this level of analysis so far (apologies
> if I’ve missed something) but I’d be keen to participate in such an
> activity.
>
> Personally I don’t see a need to mention ISO19123, WaterML2, NetCDF or any
> other specific standard at the level of this requirement, except perhaps to
> give examples of what a coverage is. The following sentence in the Charter
> does a good job of highlighting that we will look at prior art:
>
> "Where deliverables build on prior work, any variance developed by the
> Spatial Data on the Web WG will be backwards compatible with the existing
> work. The aim is to formalize existing work, not to replace or compete with
> it.”
>
> Just my 0.013p (at current exchange rates).
>
> Jon
>
>
>
>
>
> On 10 Oct 2015, at 18:44, Ed Parsons <eparsons@google.com> wrote:
>
> So would a better approach be to have less specificity in the requirement?
> Ed
>
> On Sat, 10 Oct 2015, 11:41 Peter Baumann <p.baumann@jacobs-university.de>
> wrote:
>
> yes, indeed ISO will take its time. Once there, ISO CIS will stay for many
> years as ISO's understanding of coverages.
> It will be a core decision for the SDW WG whether to bypass ISO and
> INSPIRE and establish a silo solution, or be compatible with the mainstream.
>
>
> -Peter
>
>
>
> On 2015-10-10 08:13, Simon.Cox@csiro.au wrote:
>
> Ø  ISO-19123-2 (the soon to be published ISO version of the OGC Coverage
> Implementation Schema 1.1)?
>
> ‘soon to be published’ is optimistic.
> It is not yet on the ISO/TC 211 program of work [1].
> The duration from NWIP (New Work Item Proposal) to IS (International
> Standard) is never less than 3 years, even if there is a mature starting
> document.
>
> [1] http://www.isotc211.org/pow.htm
>
> *From:* Frans Knibbe [mailto:frans.knibbe@geodan.nl
> <frans.knibbe@geodan.nl>]
> *Sent:* Friday, 9 October 2015 11:28 PM
> *To:* SDW WG Public List <public-sdw-wg@w3.org> <public-sdw-wg@w3.org>;
> Peter Baumann <p.baumann@jacobs-university.de>
> <p.baumann@jacobs-university.de>
> *Subject:* Issue 27: correction of the description of the Coverage in
> Linked Data deliverable
>
> Issue 27 <http://www.w3.org/2015/spatial/track/issues/27> is a special
> one, because it is about one of the deliverables. The Coverage in Linked
> Data deliverable <http://www.w3.org/2015/spatial/charter#cov> reads "The
> WG will develop a formal Recommendation for expressing discrete coverage
> data conformant to the ISO 19123 abstract model. ..."
>
> Peter explained that this statement probably requires some adjustment, see
>  this message
> <https://lists.w3.org/Archives/Public/public-sdw-wg/2015Apr/0024.html>,
> otherwise the deliverable will not have the proper foundation.
>
> Do I understand correctly that is is a matter of saying that the
> Recommendation will not be based on ISO-19123, but on ISO-19123-2 (the soon
> to be published ISO version of the OGC Coverage Implementation Schema
> 1.1)?
>
> We can not change the charter text, but we could add a clarification (a
> note) in the chapter about deliverables in the UCR document (Ed, Kerry or
> Phil: is that correct?).
>
> If the assumption above are correct, could someone suggest a good wording
> for the note that should be added?
>
> Regards,
> Frans
>
>