Provide useful information about the thing when
its
URI is dereferenced, using standard formats such as RDF/XML
Standard HTML practice has been to use a
parameter list, following a ‘?’, and structured as named
parameters
with assigned values separated by ‘&’. E.g., a
google search
for “stuff” brings:
(where
I have reddened the parameter
list). This is already processed nicely by all kinds of
services, but the
various services are running structured object
definitions in their own
vocabulary, as implemented by the HTTP CGI components,
ActivExen, or even
custom processes tailored to specialty representations
used in specific
applications.
Rich,
Great catch!
What does "useful" mean? Who is the beholder of such "usefulness" ?
My translations:
1. Beholder is a machine or human
2. Useful is about structured data discernible by beholder.
Thus, verifying you say, the following are bona fide entity
descriptors:
Linked Data as "a term used to describe a
recommended best practice for exposing,
sharing, and connecting pieces of data, information, and knowledge on the
Semantic Web using URIs and RDF."
No, you are describing
TimBL's meme [1] that espoused:
( bla bla, compete with data base technology
with simplicity..)
TimBL discussed how you
publish (actually inject) Linked Data into the
Web by constructing hypermedia resources in the manner
prescribed above. This
is a subtle, but extremely important point re., Linked
Data comprehension,
which starts by understanding its place in a broader
technology innovation
continuum that covers -- data access, data
representation, data integration,
and data management.
I prefer to define Linked Data as hypermedia based
structured data [2].
Meaning: you represent a calendar using structured
data representation e.g.,
using iCalendar syntax that produces a ".ics"
resource [3]
which may or may not be HTTP addressable,
referencable, or accessible. As
already outlined re. iCalendar, you could do the same
using one of the syntaxes
associated with RDF [4][5][6] e.g., HTML+RDFa,
RDF/XML, Turtle, N3, NTriples
etc... You could also do the same thing using OData
(an Atom+XML based markup
from Microsoft), GData (Atom+XML based markup from
Google), even a CSV file
(where you simply stick to 3-tuples + use of "<"
and
">" to indicate reference data i.e., NTriples-like)
etc.
2. What does this do more than a sql query
does on a database? Semantically nothing
other than the facts SQL query accesses a
database objects and this accesses web
objects or stuff..
Here what a SQL based RDBMS
won't deliver, implicitly:
1. Reference values -- most relational DBMS engines
don't support reference
values (pointers)
2. References that resolve to structured data
representation -- a Foreign Key
doesn't implicitly resolve to a relational table (or
view) that projects a
union all its Referents (primary keys and dependent
columns); of course you can
code such functionality or implement an RDBMS hybrid
e.g. Object-Relational
engine that delivers this feature via ref and deref
functions as SQL
extensions.
Here is what Linked Data adds to the mix, implicitly:
1. Reference values -- via URIs (#2 of TimBL's meme
with HTTP specificity, but
this can apply to any URI scheme if you have a custom
resolver)
2. References that resolve to structured data
representation(s) -- #3 of
TimBL's meme which is biased towards RDF as the W3C
standard syntax for
structured data representation and SPARQL as the
mechanism for implicitly
binding HTTP URI based Entity/Object Names to a
Resource Addresses (URL) that
resolves to Structured Data Representation(s).
The elegance of HTTP makes the representation of
structured data negotiable
[7].
SQL query is a data manipulation language
---? DML...
3. My question is: Is that all you want to
do with Semantic Web? If so, may be we are
done .. Time B L told us so.
Hmm..
We want to Refer to Entities/Objects across a global
InterWeb space. We want
these References to emulate pointers which have
existed since the inception of
computer programming.
When we de-reference a pointer, we want to have the
option to choose (via
negotiation) how the structured data is represented.
Our base (not sole) data
model is an Entity-Attribute-Value or
Subject-Predicate-Object graph -- which
ultimately boils down to FOL based conceptual schema.
Thus, the InterWeb (many resolvable URI schemes rather
than HTTP solely)
becomes a distributed database (that plugs in data
spaces hosting
heterogeneously shaped data). A database system that
also ultimately exploits
and demonstrates the fact that Relational Tables and
Relational Graphs all sit
atop the same conceptual schema (FOL based). A global
space where I can browse
data, realize my own limitations, and then (if need
be) send agents out on my
behalf to continue navigation and discovery, bearing
in mind its understanding
of my preferences and human limitations.
Bur
if you want to actually do something more
like processing and computing data in a
complex manner, does this allow us to do
that? Nothing I have read so far tell me
that it is capable of doing so.
SPARQL can do that, while
SPASQL (SPARQL + SQL hybrid) can do more.
When the world-view is known a SQL RDBMS (at the
current time) will trump a
pure SPARQL + RDF DBMS. There are benchmarks [8] that
have proven that
repeatedly.
When deal with reality though, where world views
manifest unpredictably and
data shape is volatile SPARQL + RDF DBMS will run
rings around a SQL DBMS.
Thus, the optimal solution (as far as I know) is a
hybrid solution, especially
one that combines RDBMS and Graph Model DBMS advances
[9][10][11].
Let me tell a story. Oracle supported Data
Manipulation language, data definition
langugae and Data control language. But all
the processing code had to be written by the
front end application, middle wear etc.
Later on they supported some event driven
processing using the concept of triggers and
stored procedures.. This is a procedural /
event driven concept implemented using data
base technology. ( a little convoluted, but
competed with procedural language
processing..)
Yes, and the end product no
matter how you cut it only works within
Oracle.
There are procedural languages that actually
allow complex computation. Java is one of
them.. Either one can explore the
concept of procedural languages that are
already in use for web development, or use
linked data concept and add on baggage later
( like oracle did..)..
Yes, and the end product no
matter what is language locked.
I think it is simplicity with sagging, heavy
baggage attached.. that will come later on..
No. Let's revisit TimBL's
meme re. hypermedia based structured data,
plus my subtle tweaks that put Linked Data back into a
palatable innovation
continuum that reflects pre Web reality and history:
1. Objects have Identity via URIs
2. URIs Resolve to Representations of their Referents
3. Data Object Representation is negotiable (so RDF
and many other approaches,
rather than RDF solely) -- HTTP facilitates this
elegantly
4. Data Representation is separate from its underlying
Conceptual Schema
5. Hypermedia Resource construction should leverage
expanse of InterWeb via
URIs when referring to related data.
1-5 will give us a dense web that that lends itself to
precision find (rather
than fuzzy search) and serendipitous discovery of
relevant things based on
individual context preferences. This is what Linked
Data truly enables IMHO.
On 10/26/10
12:38 PM, John F. Sowa wrote:
> There's an interesting new language
and system designed for secure,
> distributed computing. The
language, called Jif (Java + Information
> Flow), extends Java with
"policies", and the system is called
Fabric
> because "Fabric is more useful and
more tightly connected than webs."
>
> See below for references to Fabric,
Jif, and related articles.
> But the main point I want to make
in this note is the contrast
> between the methods for developing
Fabric and the Semantic Web:
>
> 1. The SemWeb began with an
inspiring, but rather vague speech
> by Tim B-L about adding
semantics to the URIs of the WWW.
> At that level of detail,
nobody could object.
>
> 2. The W3C, which met for the
first time at the 1994 conference
> where Tim gave that speech,
took charge of the design and
> development of the SemWeb.
>
> 3. Like any design by committee
(cf. Fred Brooks' book), the
> SemWeb was pulled in
different directions by experts with
> competing visions of the
goals, technology, and methodology.
>
> 4. As a result, the only
consensus on architectural document was
> the familiar layer cake,
which emphasized syntax over semantics.
>
> 5. The most widely used
technology that came out of the SemWeb
> was the lowest common
denominator with the barest minimum of
> semantics: RDF.
>
> 6. The components above the RDF
level have not been integrated
> with each other or with the
mainstream of IT software, and
> very few IT developers have
found any reason to use them.
>
> I don't know whether Jif and Fabric
are going to be more successful,
> but their approach is the best way
to develop a major new design:
> a small group doing focused
research with prototype implementations
> to check how and whether the ideas
work in practice.
>
> Doing advanced R& D in a small
group (or "skunk works") has always
> been far more successful than
design by committee. As a classic
> example that succeeded beyond the
designers' wildest dreams, see
> the Oak project at Sun, which
became Java:
>
> https://duke.dev.java.net/green/
>
> As Yogi B. said, "Prediction is
very hard, especially about the future."
> But I don't believe that any of the
current components of the SemWeb
> are going to survive without a
total overhaul or complete replacement.
> Instead, we can expect some small
group working in skunk-works mode
> to produce a truly Semantic Fabric.
John,
I agree with the general "skunk works"
theme 100%, but Java isn't a
great example today IMHO. Lots of bloat
in Java land (codebase and
community process).
Linked Data (hypermedia based structured
data) and the Linked Open Data
community are "skunk works" examples
that emerged from the larger and
somewhat maligned Semantic Web project
:-)