Lists Home |
Date Index |
I generally agree with you - with all due (and considerable) respect to the
academics and visionaries working on Semantic Web ideas, I think we'll be
seeing simple practical benefits of RDF long before formally-grounded
inferencing systems become a standard tool.
My own dabblings suggest to me that RDF can offer good solutions for a lot
of problems where data is largely unstructured, and where some kind of
global disambiguation is needed, such as in widely distributed systems.
Given RDF's heritage, that is really quite a good thing, because the web has
both of these characteristics. It's also notable that in many respects the
human world shares them.
The "metacrap" argument is a complete red herring, because it makes the
assumption that the creation of metadata must involve extra effort on the
part of the system/application users. In most circumstances there is stacks
of metadata on hand, and I don't have far to look for an example. Nearby
there's all the mail header date/time & routing material etc, there's a
thread in the archives. Without any extra effort on my part, there's a sig
with the addresses of my web space and some of the material I'm working on.
Linked to that there is biographical information about me. Your address is
here too, which may or may not be used to get to biographical information
about yourself, but it does describe a communication channel to you through
which more information could be obtained.
Ok, so the mail client I'm using doesn't take advantage of all this, and
that to recall any of this pile of information post-mortem scraping is
needed. This client (Outlook) is smart enough that if I didn't have defenses
then a bit of kiddy code from a third party could make it spam everyone in
my address book. Why shouldn't the client use this information in a
consistent, secure and useful fashion? C'mon mail dude, the spell checker
knows this is in UK English, why don't you?
Hopefully Chandler will.
Regarding Linda and friends, I got very excited when I first came across
Jini and JavaSpaces, but there is little evidence of widespread adoption of
what is undeniably a very nifty system. I guess a lot of the reason is down
to the demands of retooling, which though not insignificant for triplespace
is less of a burden.
Regarding the "RDF tax", one aspect of the analogy that seems to have been
overlooked is that in the real world taxes occasionally go towards
infrastructure, and at the end of the year there can be tax rebates. For
infrastructure, metadata generated for use in one system can be reused in
others. For example, there's a lot of RSS data coming from weblogs. Commonly
the items are descriptions of a web page. Why not use that information to
provide annotation for the page? Ok, to some extent we can already do that
using Google link:to, which will pick up on blogs irrespective of whether
they syndicate or not. But why choose such a sloppy approach when there is a
systematic solution just waiting to be exploited?
In vanilla XML systems like RSS 0.93, whether or not it's archived, all the
useful metadata that could be associated with the syndicated information
effectively gets pulped after publication (unless you want to go scraping
again). With RSS 1.0 the unambiguous metadata with formally defined
semantics allows for systematic cataloguing and/or retrieval. There's your
Idea maps for the Semantic Web
<stuff> http://www.isacat.net </stuff>
Semantic Web Log :
>From: Mike Champion [mailto:firstname.lastname@example.org]
>Sent: 16 November 2002 20:01
>Subject: [xml-dev] RDF for unstructured databases, RDF for axiomatic
>systems ?(was Re: [xml-dev] XML/RDF )
>11/14/2002 4:20:30 PM, Uche Ogbuji <email@example.com> wrote:
>>I should say that the project is proving a resounding success
>within Sun. So
>>much for the idea that RDF is some academic oddity.
>I think I am beginning to grok the RDF world a little better. Perhaps
>the problem (at least the motivation for the assertions/questions
>Uche's scorn) was that I had seen RDF through the lens provided by
>effort and the Semantic Web vision, that is, as a way to define
>on which machines will make interesting logical inferences about
>the web. For example, in the TimBL et al Sci Am article:
>"Adding logic to the Web?the means to use rules to make inferences,
>choose courses of action and answer questions?is the task before the
>Semantic Web community at the moment. A mixture of mathematical and
>engineering decisions complicate this task. The logic must be powerful
>enough to describe complex properties of objects but not so powerful
>that agents can be tricked by being asked to consider a paradox...
>An important facet of agents' functioning will be the exchange of "proofs"
>written in [RDF]".
>That strikes me as highly unlikely to evolve, because a) it's
>to build consistent and powerful axiomatic systems, especially about human
>affairs; b) logical inferences are extremely fragile in the face
>of ambiguity and
>inconsistency, and the "solution" seems to presume profound breathroughs in
>logic and/or computer science, and c) the "metacrap" issues: it's
>very clear that people
>(stupid, lazy, greedy, un-knowing of ourselves that we are)
>will not take the trouble to provide good and honest metadata for
>unless The Boss is watching closely. Not to mention the fact that
>few (e.g. the
>RSS users) want to pay the RDF tax, given its XML serialization
>that seems to
>have no friends at all except the RDF working group charter :-)
>BUT perhaps I have been oblivious to the REAL users of RDF (and other
>semantic mapping technologies) who seem to use rough 'n ready
>"<street>, <rue>, and <strasse> can be considered synonymous in an
>context"), and those who use it as sortof an unstructured database
>management applications. If one thinks of RDF queries as following
>"reasoning" that can ignore inconsistencies (I think Uche mentioned the
>heuristic of using the first assertion) rather than as rigorous proofs
>in a logico-deductive system, exploiting RDF's recursive
>structure common to most [all?] natural languages, then maybe some
>interesting things can happen.
>I'm still unpersuaded that very many HAVE happened. We can all
>claim that our favorite technology is an "academic oddity" by reference to
>a handful of proofs-of-concepts along with the assertion that this is the
>tip of the iceberg or the start of something big. (My favorite "academic
>oddity" Next Big Thing candidate at the moment is Linda / tuple spaces /
>XML spaces ... it would be an interesting exercise to see if I could
>come up with as many success stories for this as an RDF stakeholder
>could for RDF ... <grin> ).
>Still, I wouldn't bet against a simplified RDF syntax
>(XML or otherwise) latching onto the "unstructured database" meme
>and growing into something Really Big. That will end up
>looking about as much like the Semantic Web vision as the
>HTTP/HTML web looks like
>Ted Nelson's vision ... But what the hell, nobody ever said that evolution
>favors the the most beautiful, only that it favors practical solutions to
>The xml-dev list is sponsored by XML.org <http://www.xml.org>, an
>initiative of OASIS <http://www.oasis-open.org>
>The list archives are at http://lists.xml.org/archives/xml-dev/
>To subscribe or unsubscribe from this list use the subscription