On 5/27/11 3:03 AM, John F. Sowa wrote:
> On 5/26/2011 3:11 PM, Kingsley Idehen wrote:
>> ... but today you have middleware solutions that produce
>> Linked Data Views over relational data sources. There's even an R2RML
>> [1] effort from the W3C to standardize this effort. That said, I grasp
>> the context of your gripe. This should have been one of the first areas
>> of total emphasis back in the day.
> But those middleware solutions are *trivial* compared to the research
> that was done in the 1990s. (01)
Not all of them. I wouldn't consider what we do in Virtuoso (which by
the way is built by a team of developers with strong LISP backgrounds
amongst other things) as being trivial. We are able to make scalable
Linked Data views over relational data sources on the fly etc.. We have
a hybrid data server (which combines conventional database management
and middleware functionality). (02)
Virtuoso came to be in 1998. (03)
> The SW technology available in 2011 could
> have been developed far better a decade ago if the W3C gang had done
> their homework and read the published research in the 1990s. (04)
Of course, but we can't put every technology that supports output from
the SW technology project as being the same. One shoe doesn't fit all. (05)
> Please read the following article from 1994 -- the same year that
> Tim B-L gave his first speech on his vision for the Semantic Web:
>
> http://wwwis.win.tue.nl/~gfletcher/2id35-spring11/reading/tsatalos.pdf
> A Unified Framework for Indexing in Database Systems (06)
Please do understand that when it comes to DBMS and Middleware matters I
don't take my cues from TimBL or the W3C, I just don't. I come from a
different place re. data virtualization, access, integration, and
management. (07)
> The authors presented a more detailed version at the VLDB conference
> in 1996:
>
> http://tinyurl.com/3pedzwd
> The GMAP: A Versatile Tool for Physical Data Independence
>
>> It took a long time for SPARQL to emerge from the W3C's Semantic Web
>> project, relative to the inaugural article in Scientific American.
> But what you really need is the option of executing both SPARQL
> queries and SQL queries on *exactly the same data* without requiring
> any mapping or reformatting -- and that's what you get with Gmaps
> (Generalized Combinatorial Maps). (08)
Yes, and that's exactly what we do in Virtuoso which supports SQL,
SPASQL (SPARQL inside SQL), PSM-96 Persistent Stored Modules (SQL Stored
Procedures), 3rd Party Runtime hosting for User Defined Types etc.. . Do
note, that we simply leverage W3C standards as part of our fundamental
commitment to standards based interop. (09)
BTW - we have a declarative meta schema language, and the only value
that R2RML offers to us is the ability to import and export declarations
of Linked Data Views that's portable with other systems to do similar.
> The authors demonstrated that the Gmap encoding could support *both*
> relational queries (e.g. SQL) and path-based queries (similar to SPARQL)
> on exactly the same Gmap representations. Furthermore, the Gmap format
> required *fewer* disk accesses than the native relational storage for
> SQL queries or the native OODB storage for path-based queries.
>
> That is the best of both worlds -- and they implemented it in 1994. (010)
Hmm. It was just OK :-) (011)
We need to be able to Reference Data Objects across machine boundaries,
GMAP doesn't do that (at least when I looked way back). By leveraging
whatever standards exist, interop is possible. Thus, in the case of W3C
standards (or others where such exist) you can go further, and that's
where de-referencable URIs based Linked Data Graphs are vital. (012)
As you said, URIs and HTTP applied to graph structures is powerful, it
doesn't need to be encumbered by syntax, which is where things are
getting problematic re. Linked Data and the W3C, courtesy of RDF.
> The architects have some of the largest applications of Gmaps for
> storing and accessing huge graphs. A shopping mall with multiple
> stores connected to a couple of large hotels, a movie theater,
> etc., can be represented by a graph with billions of nodes. (013)
Yes, we have a 23 Billion+ live instance of Virtuoso [1] that's been
online for month. I can even handle complex inference rules at that scale.
> And Gmaps are so efficient that they can dynamically create virtual
> reality views in real time. They can let you "walk" through the
> mall at any speed and look around as you go. They even let you
> take a virtual helicopter flight over the buildings and recreate
> the views from any viewing angle. (014)
Yes. (015)
This is only exciting to me if they can pull this off with fewer
computer resources that we do, and more importantly, make data access
available to anyone by URI style Name References. In addition, a live
endpoint for ad-hoc queries that allows anyone perform aggregates and
joins etc. I would very much enjoy evaluating this system. (016)
> For more info, use your favorite search engine to find references.
> But the acronym Gmaps is ambiguous, since it can also mean "Google
> maps." So add the word 'combinatorial' to your search list.
>
>> I really believe that via Linked Data Graphs the critical
>> foundation opens up for much broader participation that truly
>> works the broader tech continuum. It's going to happen since
>> you can never really hold a good thing down!
> I agree, but a group with a lot of influence, such as the W3C, can
> either promote technology or hold it back. In this case, it's been
> 17 years since that publication in 1994, and the W3C still hasn't
> got the message. (017)
Sorta, I still have my fingers crossed re. their ability to learn from
mistakes made in the past. (018)
Links showcasing various Virtuoso capabilities re. SPARQL (all of which
you can execute from inside SQL and JOIN against Tables and Procedure
Views etc.): (019)
1. http://lod.openlinksw.com -- live Virtuoso instance (note this
product is a multi-model DBMS engine that supports a variety of industry
standards re. data access and management) (020)
2. http://lod.openlinksw.com/sparql -- SPARQL endpoint (021)
3. http://lod.openlinksw.com/b3s/ -- collection of sample SPARQL Queries
(if a query doesn't give a result you can cut and paste the sparql then
apply to one of the endpoints) (022)
4.
http://lod.openlinksw.com/describe/?url=http%3A%2F%2Fdata.ordnancesurvey.co.uk%2Fid%2F7000000000000018 (023)
-- page demonstrating faceted navigation and use of reasoning to
determine labels for TBox and ABox data (so we also have a best effort
heuristic for determining labels across inconsistent data so as to
reduce surfacing of raw URIs in anchor text) (024)
5.
http://lod.openlinksw.com/describe/?url=http://www.mpii.de/yago/resource/wordnet%255Fmonarchy%255F108363812 (025)
-- looking at information about Monarchy (026)
6. http://lod.openlinksw.com/c/CO6X3YX -- places of worship
(denomination unspecified) within close proximity of cafes in Paris (027)
7. http://www.delicious.com/kidehen/linked_data_demo -- vast collection
of demos (many use the massive LOD cloud cache instance others use the
smaller URIBurner instance). (028)
Kingsley
> John
>
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
>
> (029)
-- (030)
Regards, (031)
Kingsley Idehen
President& CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen (032)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (033)
|