RDF: the emporer's new clothes

Discussion in 'XML' started by sci-ontologist, Oct 25, 2006.

  1. This is not meant to be a troll. I'm not at all sure I believe what
    follows. I *might* believe it.
    And I am curious to see what reaction it will get from this group:

    ======
    RDF is a technology for attaching "semantic meaning"
    to "web resources," and also a way to advertise that meaning.

    Figuring out "semantic meaning" amounts to a natural language
    problem. This means it cannot be done (or done very well) by
    computers. So, if RDF cannot be generated robotically,
    and if useful amounts of queriable RDF data are ever
    going to exist, then that RDF data will have to be manually
    created, essentially from the keyboard.

    Further, because RDF has an obscure, user-unfriendly syntax,
    we can assume it cannot and will not be hand edited by
    humans who are NOT "knowledge engineers."
    In other words, only highly trained smart people can ever be
    expected to make RDF.

    Finally, because useful RDF data can only
    be produced by highly trained professionals, it will in fact,
    never get produced. So this whole RDF-semantic-web enterprise
    is a fantasy that will never happen....or, at best, it will
    exist as small scale pockets of locally and expertly produced
    data.
    sci-ontologist, Oct 25, 2006
    #1
    1. Advertising

  2. sci-ontologist

    Andy Dingley Guest

    sci-ontologist wrote:
    > RDF is a technology for attaching "semantic meaning"
    > to "web resources," and also a way to advertise that meaning.


    No, that's just one application of it. It's not about "metadata" either
    (altough that's a popular application for it too).

    If you like, think of RDF as nothing more than "XML 2.0"

    > Figuring out "semantic meaning" amounts to a natural language
    > problem.


    Yes. It's hard. That's why one of the first principles of metadata is
    "Catch it early", before you have to start inferring it.
    (was it Carl Lagoze who codified these principles?)

    The admitted difficulty of inferring semantic content from a
    pre-existing but undescribed resource is a problem -- but it doesn't
    mean you have to throw the bayb out with the bathwater and abandon all
    attempts, including tools like RDF.

    > This means it cannot be done (or done very well) by
    > computers. So, if RDF cannot be generated robotically,


    There's your first trollish non-sequitur for you.

    > and if useful amounts of queriable RDF data are ever
    > going to exist,


    They've existed for some year. I've _personally_ generated at least a
    gigabyte of it, data which someone else pays hard cash money to query.

    > then that RDF data will have to be manually
    > created, essentially from the keyboard.


    By that same logic the Word .doc file format is equally doomed.


    > Further, because RDF has an obscure, user-unfriendly syntax,


    RDF doesn't have a syntax
    RDF/XML has a syntax which had the goals of expressing RDF and being in
    XML first, readability second. If you want a more readable syntax, use
    N3 triples


    > we can assume it cannot and will not be hand edited by
    > humans who are NOT "knowledge engineers."


    Who ever claimed that knowledge engineers or content creators would
    have to hand-edit anything?
    Andy Dingley, Oct 25, 2006
    #2
    1. Advertising

  3. Andy Dingley wrote:

    > They've existed for some year. I've _personally_ generated at least a
    > gigabyte of it, data which someone else pays hard cash money to query.


    ....exactly my point. If you want to have RDF you have to hire someone
    like Andy Dingley to get it.

    There is a lot of (government funded) interest in RDF as a way to make
    public domain
    scientific data available, in a 'platform independent way.'

    Most of that data is (currently) stored in myriad disparately defined
    relational schemas, stored on architecturally disparate servers
    (Oracle, Informix, MySql, etc). There is a lot of interest in figuring
    out how
    to make an alternate, platform independent RDF "view" of that
    already-publicly-funded data.

    I don't think it will ever happen. There is no way to read
    an arbitrary data dictionary and then spit out RDF.
    That job has to be custom designed, on a schema by schema
    basis, by a programmer.

    Annotating data once is expensive (inserting it into a relational
    schema).
    Annotating it a second time (turn it into RDF) is even more expensive.
    That was my point. Sure, RDF is possible. But it is and will always
    be extra-expensive to produce.
    sci-ontologist, Oct 25, 2006
    #3
  4. sci-ontologist

    Nick Kew Guest

    In article <>,
    "Andy Dingley" <> writes:

    > If you like, think of RDF as nothing more than "XML 2.0"


    YOUR WIFE IS A BIG HIPPO[1]!

    > By that same logic the Word .doc file format is equally doomed.


    wishful thinking ...

    [1] google is your friend, if you don't recognise the reference.

    --
    Nick Kew

    Application Development with Apache - the Apache Modules Book
    http://www.apachetutor.org/
    Nick Kew, Oct 25, 2006
    #4
  5. sci-ontologist, Oct 25, 2006
    #5
  6. sci-ontologist

    Andy Dingley Guest

    Nick Kew wrote:

    > YOUR WIFE IS A BIG HIPPO[1]!


    Sounds like too much risk of pterry references, IMHO....

    I've never understood how people could do anything useful over IRC.
    Hateful thing. If I could handle IRC, I'd go outdoors and talk to
    people.
    Andy Dingley, Oct 25, 2006
    #6
  7. sci-ontologist

    Andy Dingley Guest

    sci-ontologist wrote:

    > Andy Dingley wrote:
    >
    > > They've existed for some year. I've _personally_ generated at least a
    > > gigabyte of it, data which someone else pays hard cash money to query.

    >
    > ...exactly my point. If you want to have RDF you have to hire someone
    > like Andy Dingley to get it.


    Not at all. You can hire me to write Word macros (and better things,
    but Word met one customer's needs) which let your cheap staff churn out
    RDF documents with minimal training.

    My current project's build and release documentation runs through about
    1MB of RDF a day. A bunch of nasty scripts (shell, Perl, Python and
    damp string) make RDF/XML/ Dublin Core documents, then they're
    processed by XSLT into the status and release documentation for about
    two dozen product streams. Not only does this work, but I didn't even
    have to do much work to make it fly. I already have a large toolset of
    RDF / DC tools, so the "abstract document to readable publication" step
    is a very easy part of the dev process (when I grow up I want to be
    Norman Walsh)

    Firefox is held together internally with RDF.

    The Protege crowd are doing things far smarter than I can even imagine.

    RDF is already _there_.


    > There is a lot of (government funded) interest in RDF as a way to make
    > public domain
    > scientific data available, in a 'platform independent way.'


    Have you noticed how well it works too? (So long as you're
    Australian, not the crappy UK eGMS efforts)

    > Most of that data is (currently) stored in myriad disparately defined
    > relational schemas, stored on architecturally disparate servers
    > (Oracle, Informix, MySql, etc). There is a lot of interest in figuring
    > out how
    > to make an alternate, platform independent RDF "view" of that
    > already-publicly-funded data.


    Damn right. And it's happening.


    > I don't think it will ever happen. There is no way to read
    > an arbitrary data dictionary and then spit out RDF.
    > That job has to be custom designed, on a schema by schema
    > basis, by a programmer.


    Firstly, so what ? Even this task is no worse than plenty of pre-RDF
    interchange techniques. I do this a lot. I have a choice whether I use
    RDF or not, but don't have a choice whether I connect the systems or
    not - that's my job. I've been doing it without RDF for years, but this
    way is quicker, easier and more re-usable.

    Secondly, you're largely wrong. If you have a formal data model (such
    as much decently architected SQL) then RDF-out publication is _very_
    easy. It's going the other way that's harder.

    > Annotating data once is expensive (inserting it into a relational
    > schema).


    Annotating data can be cheap, so long as you do it at the same time as
    data creation and with the same tools and user involvement. The closer
    you couple these, the cheaper they get. In many cases this can become a
    negligible addition (if capture is simply automatic).

    > Annotating it a second time (turn it into RDF) is even more expensive.


    "Annotating" isn't the same thing as "turning it into" something.
    Annotating is certainly expensive - generally the cost of human
    data-entry, the technology is negligible. Transcoding though from a
    specific domain to a general is simple and cheap. If you already have
    annotation by some arcane, task-specific but competent process, then
    getting it out is cheap.
    Andy Dingley, Oct 25, 2006
    #7
  8. RE> RDF is already there.......

    Well I'll bow out now and think about all this.
    I did say I wasn't sure I believed any of what I was
    about to say..........that I wanted to hear someone
    else's response. I got that, and some.

    It is clear to me that RDF is not going to replace
    databases. So, if RDF is going to be an important,
    alternate way to find and query data, all new
    DB applications will have to be designed to
    still respond to SQL (or xpath/xquery) and also
    to know how to produce RDF.

    So now there will be at least two ways to query
    the data. We will, at the very least, have to learn
    how to walk and chew gum at the same time.
    sci-ontologist, Oct 26, 2006
    #8
  9. > It is clear to me that RDF is not going to replace
    > databases.


    It was never intended to do so.

    > So now there will be at least two ways to query
    > the data.


    There are a near-infinite number of ways to query data; XQuery isn't a
    complete solution either, nor is SQL. The question is which ones are
    useful for specific tasks.

    Use what solves your problems and those of your customers and their
    customers; ignore what doesn't. And remember that using XML as your
    interchange format does not necessarily make it the best representation
    in the back end of the system.

    Don't get caught up in absolutes; they are always inherently false.
    Including this one.

    --
    () ASCII Ribbon Campaign | Joe Kesselman
    /\ Stamp out HTML e-mail! | System architexture and kinetic poetry
    Joe Kesselman, Oct 26, 2006
    #9
  10. sci-ontologist

    Andy Dingley Guest

    sci-ontologist wrote:

    > It is clear to me that RDF is not going to replace
    > databases.


    Who ever claimed it would do? Saying that it will is like claiming
    that photocopying replaces postage or filing. One's a document, one's a
    process for storage or retrieval.

    There are RDF databases certainly. SPARQL is a query language for them
    and I've done work myself on implementing powerful large volume
    databases to store the stuff. These aren't "RDF" though, they're just
    something else that uses it.

    As to whether "RDF databases" will replace "relational databases", then
    I doubt if they ever will in general, but they already have begun to
    for some problems with semi-structured data (data that has structure,
    but that doesn't advertise this structure beforehand, such that you can
    set up a data model of it before beginning to store it). Just look at
    the recent Oracle work on graph-model storage.

    Personally I don't see XML databases as having a bright future.
    Relational is good (we all have utility bills of much the same
    structure) and there's a developing need for the semi-structured case.
    XML though falls in the middle -- you need the pre-defined schema, but
    hierarchical tree isn't much better than relational for storage of
    real-world graph data.


    > So, if RDF is going to be an important,
    > alternate way to find and query data, all new
    > DB applications will have to be designed to
    > still respond to SQL (or xpath/xquery) and also
    > to know how to produce RDF.


    I'm not even going to bother refuting that one. You can produce
    pointless straw men all day.
    Andy Dingley, Oct 26, 2006
    #10
  11. Andy Dingley wrote:
    > > So, if RDF is going to be an important,
    > > alternate way to find and query data, all new
    > > DB applications will have to be designed to
    > > still respond to SQL (or xpath/xquery) and also
    > > to know how to produce RDF.

    >
    > I'm not even going to bother refuting that one. You can produce
    > pointless straw men all day.


    Refute? I said I was going to bow out and think about
    all this. But this point is important, at least to me. Government
    funded
    research (grant money) in the US is awarded by peer
    reviewers who (in the IT/informatics context)
    are now beginning to turn down grant requests--
    for database related projects--if they don't promise
    to "support RDF"

    So, in that political sense, for better or for worse,
    RDF is now being injected into the DB/Informatics realm.
    Weather you like it or not. Not everybody is a such
    a big fan:

    "Tim Bray responded "I'd go further. I think the current RDF/XML
    syntax is so B.A.D. (broken as designed) that it has seriously got in
    the way of people being open-minded about RDF. I'm baffled why the
    RDF working group has been forbidden to work on replacing that syntax."
    sci-ontologist, Oct 26, 2006
    #11
  12. sci-ontologist

    Andy Dingley Guest

    sci-ontologist wrote:

    > Government
    > funded
    > research (grant money) in the US is awarded by peer
    > reviewers who (in the IT/informatics context)
    > are now beginning to turn down grant requests--
    > for database related projects--if they don't promise
    > to "support RDF"


    Cite the evidence please.


    > "Tim Bray responded "I'd go further. I think the current RDF/XML
    > syntax is so B.A.D. (broken as designed) that it has seriously got in
    > the way of people being open-minded about RDF. I'm baffled why the
    > RDF working group has been forbidden to work on replacing that syntax."


    This is such a misquotation as to be positively deceitful.

    Firstly you keep confusing RDF and RDF/XML

    Secondly, Tim Bray said this in 2002 when the RDF/XML syntax _was_
    ghastly and the RDF WG were busy on documenting and replacing it -- a
    mammoth (and excellent) piece of work which culminated in the whole RDF
    documentation family being replaced in early 2004. I have no idea why
    he thought they were "forbidden" from doing this, or even who had the
    power to forbid it. The task wasn't specifically to do that, but they
    certainly did clarify it,

    Nor was the original RDF/XML even particularly bad in the simple cases.
    The problem was that it was unusably unclear in the non-simple cases.
    This got fixed in 2004.
    Andy Dingley, Oct 26, 2006
    #12
  13. 1)
    Take a deep breath, and try have a conversation
    without so easily losing your cool. I'm easy.
    You could be too.

    2)
    RE> "Tim Bray taken out of context........"
    I don't think so. I don't know Tim Bray, but If you read the entire
    thread surrounding
    that discussion (what's wrong with RDF at xml.com) you'll see
    Tim Bray complained that unlike HTML, which was simple
    enough for ordinary mortals to learn and use, RDF
    was too complex for widespread use.

    That doesn't sound like a minor syntax tweak that got
    fixed in 2004.
    sci-ontologist, Oct 26, 2006
    #13
  14. sci-ontologist

    Andy Dingley Guest

    sci-ontologist wrote:

    > Tim Bray complained that unlike HTML, which was simple
    > enough for ordinary mortals to learn and use, RDF
    > was too complex for widespread use.


    If HTML (which is SGML) was "simple enough for ordinary mortals", then
    why did Tim Bray need to work on XML as an accessible and usably
    simplified development of SGML? As any competent HTML coder (these are
    rare and found almost only in c.i.w.a.h) knows, HTML is anything _but_
    simple in its subtleties, nor are "ordinary mortals" attempts at coding
    it anywhere near valid.

    RDF/XML is verbose and unfriendly. Fortunately I don't write it myself,
    I have computers to do that for me. What's the problem?

    If you have a pressing need to hand-author, then look at RelaxNG and N3
    triples.
    Andy Dingley, Oct 27, 2006
    #14
  15. In article <>,
    Andy Dingley <> wrote:

    >If HTML (which is SGML) was "simple enough for ordinary mortals", then
    >why did Tim Bray need to work on XML as an accessible and usably
    >simplified development of SGML?


    That HTML is simple does not imply SGML is simple, any more than XML
    being simple does.

    -- Richard
    Richard Tobin, Oct 27, 2006
    #15
  16. sci-ontologist

    Andy Dingley Guest

    Richard Tobin wrote:

    > That HTML is simple does not imply SGML is simple, any more than XML
    > being simple does.


    That was a bit cheeky of me :cool:

    Still, HTML is far from simple to get right! (and it usually isn't).
    Although XML might be simple enough for humans, and RDF/XML might be
    harder, then I don't see RDF/XML as being significantly or unworkably
    any more dificult than non-XHTML HTML
    Andy Dingley, Oct 27, 2006
    #16
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. martin

    RDF reading in .NET

    martin, Aug 4, 2003, in forum: ASP .Net
    Replies:
    1
    Views:
    447
    Steven Livingstone
    Aug 4, 2003
  2. Toby A Inkster

    Re: What's RSS/RDF etc.?

    Toby A Inkster, Jul 7, 2003, in forum: HTML
    Replies:
    1
    Views:
    397
    spaghetti
    Jul 7, 2003
  3. mj002c7991

    Re: What's RSS/RDF etc.?

    mj002c7991, Jul 8, 2003, in forum: HTML
    Replies:
    0
    Views:
    418
    mj002c7991
    Jul 8, 2003
  4. Peter Saffrey

    RDF/XLink for linkage in XML Schema

    Peter Saffrey, Jul 1, 2003, in forum: XML
    Replies:
    0
    Views:
    870
    Peter Saffrey
    Jul 1, 2003
  5. hhf77

    sell clothes

    hhf77, Feb 11, 2008, in forum: C++
    Replies:
    0
    Views:
    438
    hhf77
    Feb 11, 2008
Loading...

Share This Page