RDF-XML.html 16.5 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
<html xmlns="http://www.w3.org/1999/xhtml">
  <head>
    <meta name="generator" content=
    "HTML Tidy for Mac OS X (vers 31 October 2006 - Apple Inc. build 13), see www.w3.org" />
    <meta http-equiv="Content-Type" content="text/html" />
    <title>
      Semantic Web: Why RDF is more than XML
    </title>
    <link href="di.css" rel="stylesheet" type="text/css" />
  </head>
  <body bgcolor="#DDFFDD" text="#000000">
    <address>
      Tim Berners-Lee
      <p>
        <small>Date: September 1998. Last modified: $Date:
        1998/10/14 20:17:13 $</small>
      </p>
      <p>
        Status: An attempt to explain the difference between the
        XML and RDF models. Editing status: Draft. Comments
        welcome!
      </p>
    </address>
    <p>
      <a href="Overview.html">Up to Design Issues</a>
    </p>
    <hr />
    <h1>
      Why RDF model is different from the XML model
    </h1>
    <p>
      This note is an attempt to answer the question, "Why should I
      use RDF - why not just XML?". This has been a question which
      has been around ever since RDF started. At the W3C Query
      Language workshop, there was a clear difference of view
      between those who wanted to query documents and those who
      wanted to extract the "meaning" in some form and query that.
      This is typical. I wrote this note in a frustrated attempt to
      explain whatthe RDF model was for those who though in terms
      of the XML model. I later listened to those who thought in
      terms of the XML model, and tried to writ it the other way
      around in <a href="XML-Semantics.html">another note</a>. This
      note assumes that the XML data model in all its complexity,
      and the RDF syntax as in RDF Model and Syntax, in all its
      complexity. It doesn't try to map one directly onto the other
      -- it expresses the RDF model using XML.
    </p>
    <p>
      Let me take as an example a single RDF assertion. Let's try
      "The author of the <i>page</i> is <i>Ora</i>". This is
      traditional. In RDF this is a triple
    </p>
    <pre>
triple(author, page, Ora)
</pre>
    <p>
      which you can think of as represented by the diagram
    </p>
    <p align="center">
      <img src="diagrams/aac.gif" width="265" height="73" alt=
      "page ---has author---&gt; Ora" border="0" />
    </p>
    <p>
      How would this information be typically be represented in
      XML?
    </p>
    <pre>
&lt;author&gt;
     &lt;uri&gt;page&lt;/uri&gt;
     &lt;name&gt;Ora&lt;/name&gt;
&lt;/author&gt;
</pre>
    <p>
      or maybe
    </p>
    <pre>
&lt;document href="page"&gt;
   &lt;author&gt;Ora&lt;/author&gt;
&lt;/document&gt;
</pre>
    <p>
      or maybe
    </p>
    <pre>
&lt;document&gt;
   &lt;details&gt;
    &lt;uri&gt;href="page"&lt;/uri&gt;
    &lt;author&gt;
        &lt;name&gt;Ora&lt;/name&gt;
    &lt;/author&gt;
    &lt;/details&gt;
&lt;/document&gt;
</pre>
    <p>
      or maybe
    </p>
    <pre>
&lt;document&gt;
   &lt;author&gt;
    &lt;uri&gt;href="page"&lt;/uri&gt;
    &lt;details&gt;
        &lt;name&gt;Ora&lt;/name&gt;
    &lt;/details&gt;
    &lt;/author&gt;
&lt;/document&gt;

&lt;document href="http://www.w3.org/test/page" author="Ora" /&gt;
</pre>
    <h2>
      The XML Graph
    </h2>
    <p>
      These are all perfectly good XML documents - and to a person
      reading then they mean the same thing. To a machine parsing
      them, they produce different XML trees. Suppose you look at
      the XML tree
    </p>
    <pre>
&lt;v&gt;
   &lt;x&gt;
    &lt;y&gt; a="ppppp"&lt;/y&gt;
    &lt;z&gt;
        &lt;w&gt;qqqqq&lt;/w&gt;
    &lt;/z&gt;
   &lt;/x&gt;
&lt;/v&gt;
</pre>
    <p>
      It's not so obvious what to make of it. The element names
      were a big hint for a human reader.
    </p>
    <p>
      <b>Without looking at the schema</b>, you know things about
      the document structure, but nothing else. You can't tell what
      to deduce. You don't know whether <i>ppppp</i> is a <i>y</i>
      of <i>qqqqq</i>, or <i>qqqqq</i> is a <i>z</i> of
      <i>ppppp</i> or what. You can't even really tell what real
      questions can be asked. A source of some confusion is that in
      the xyz example above, there are lots of questions you
      <i>can</i> ask. They are questions like,
    </p>
    <ul>
      <li>Is there a w element within a details element?
      </li>
      <li>What is the content of the w element within the first x
      element?
      </li>
      <li>What is the content of the w element following the first
      y element which contains an x element whose a attribute is
      "pppp"?
      </li>
      <li>and so on.
      </li>
    </ul>
    <p>
      These are all questions about the <i>document</i>. If you
      know the document schema (a big <i>if</i>) , and if that
      schema it only gives you a limited number of ways of
      expressing the same thing (another big <i>if</i>) , then
      asking these questions can be in fact equivalent to asking
      questions like
    </p>
    <ul>
      <li>What is the author of <i>page</i>?
      </li>
    </ul>
    <p>
      This is hairy. It is possible because there is a mapping from
      XML documents to semantic graphs. In brief, it is hairy
      because
    </p>
    <ul>
      <li>The mapping is many to one
      </li>
      <li>You need a schema to know what the mapping is
      </li>
      <li>(The schemas we are talking about for XML at the moment
      do not include that anyway and would have to have a whole
      inference language added)
      </li>
      <li>The expression you need for querying something in terms
      of the XML tree is necessarily more complicated than the
      expression you need for querying something in terms of the
      RDF tree.
      </li>
    </ul>
    <p>
      This last is a big one. If you try to write down the
      expression for the author of a document where the information
      is in some arbitrary XML schema, you can probably do it
      though it may or may not be very pretty. If you try to
      combine more than one property into a combined expression,
      (give me a list of books by the same author as this one),
      saying it in XML gets too clumsy to consider.
    </p>
    <p>
      (Think of trying to define the addition of numbers by regular
      expression operations on the strings. Its possible for
      addition. When you get to multiplication it gets ridiculous -
      to solve the problem you would end up reinventing numbers as
      a separate type.)
    </p>
    <p>
      Looking at the simple XML encoding above,
    </p>
    <pre>
&lt;author&gt;
     &lt;uri&gt;page&lt;/uri&gt;
     &lt;name&gt;Ora&lt;/name&gt;
&lt;/author&gt;
</pre>
    <p>
      it could be represented as a graph
    </p>
    <p>
      <img src="diagrams/xml1.gif" alt=
      "A graph of the XML tree with 3 element nodes each with name and some with content"
      width="" height="0" />
    </p>
    <p>
      We can represent the tree more concisely if we make a
      shorthand by writing the name of each element inside its
      circle:
    </p>
    <p>
      <img src="diagrams/aab.gif" width="" height="0" />
    </p>
    <p>
      Of course the RDF tree which this represents (although it
      isn't obvious from the XML tree except to those who know) is
    </p>
    <p align="center">
      <img src="diagrams/aac.gif" width="265" height="73" alt=
      "page ---has author---&gt; Ora" border="0" />
    </p>
    <p>
      Here we have made a shorthand again by putting making the
      label for each part its URI.
    </p>
    <p>
      The complexity of querying the XML tree is because there are
      in general a large number of ways in which the XML maps onto
      the logical tree, and the query you write has to be
      independent of the choice of them. So much of the query is an
      attempt to basically convert the set of all possible
      representations of a fact into one statement. This is just
      what RDF does. It gives you some standard ways of writing
      statements so that however it occurs in a document, they
      produce the same effect in RDF terms. The same RDF tree
      results from many XML trees.
    </p>
    <p>
      Wouldn't it be nice if we could label our XML so that when
      the parser read it, it could find the assertions (triples)
      and distinguish their subjects and objects, so as to just
      deduce the logical assertions without needing RDF? This is
      just what RDF does, though.
    </p>
    <h2>
      The RDF Graph
    </h2>
    <p>
      In fact RDF is very flexible - it can represent this triple
      in many ways in XML so as to be able to fit in with
      particular applications, but just to pick one way, you could
      write the above as
    </p>
    <pre>
&lt;Description about="http://www.w3.org/test/page" Author ="Ora" /&gt;
</pre>
    <p>
      I have missed out the stuff about namespaces. In fact as
      anyone can create or own the verbs, subjects and objects in a
      distributed Web, any term has to be identified by a URI
      somehow. This actual real example works out to in real life
      more like
    </p>
    <pre>
&lt;?xml version="1.0"?&gt;
  &lt;Description

          xmlns="http://www.w3.org/TR/WD-rdf-syntax#"
          xmlns:s="http://docs.r.us.com/bibliography-info/"
 
                  about="http://www.w3.org/test/page" 
                  s:Author ="http://www.w3.org/staff/Ora" /&gt;
</pre>
    <p>
      You can think that the "description" RDF element gives the
      clue to the parser as to how to find the subjects, objects
      and verbs in what follows.
    </p>
    <p>
      This is pretty much the most shorthand way of using the base
      RDF in XML. There are others which are longer, but more
      efficient when you have, for instance, sets of many
      properties of the same object. The useful thing is that of
      course they all convey the same triple
    </p>
    <p align="center">
      <img src="diagrams/aac.gif" width="265" height="73" alt=
      "page ---has author---&gt; Ora" border="0" />
    </p>
    <p>
      It is a mess when you use questions about a document to try
      to ask questions about what the document is trying to convey.
      It will work. In a way. But flagging the grammar explicitly
      (RDF syntax is a way of doing this) is a whole lot better.
    </p>
    <p>
      Things you can do with RDF which you can't do with XML
      include
    </p>
    <ul>
      <li>You can parse the semantic tree, which end up giving you
      a set of (possibly mutually referential) triples and then you
      can use the ones you want ignoring the ones you don't
      understand.
      </li>
    </ul>
    <p>
      Problems with basing you understanding on the structure
      include
    </p>
    <ul>
      <li>Without having gone to the trouble of getting the schema,
      or having an application hand-programmed to recognise a
      particular document type, you can't pick up any semantic
      information from a document;
      </li>
      <li>When an XML schema changes, it could typically introduce
      new intermediate elements (like "details" in the tree above
      or "div" is HTML). These may or may or may not invalidate any
      query which has been based on the structure of the document.
      </li>
      <li>If you haven't gone to the trouble of making a semantic
      model, then you may not have a well defined one.
      </li>
    </ul>
    <p>
      I'll end this with some examples of the last problem. Clearly
      they can be avoided by good design even in an XML system
      which does not use RDF. Using RDF makes things easier.
    </p>
    <h2>
      Get it right
    </h2>
    <p>
      If you haven't gone to the trouble of making a semantic
      model, then you may not have a well defined one. What does
      that mean? I can give some general examples of ambiguities
      which crop up in practice. In RDF, you need a good idea about
      what is being said about what, and they would tend not to
      arise.
    </p>
    <p>
      Look at a label on the jam jar which says: "Expires 1999".
      What expires: the label, or the jam? Here the ambiguity is
      between a statement about a statement about a document, and a
      statement about a document.
    </p>
    <p>
      Another example is an element which qualifies another
      apparently element. When information is assembled in a set of
      independently thrown in records often ambiguities can arise
      because of the lack of logic. HTTP headers (or email headers)
      are a good example. These things can work when one program
      handles all the records, but when you start mixing records
      you get trouble. In XML it is all too easy to fall into the
      trap of having two elements, one describing the author, and a
      separate one as a flag that the "author" element in fact
      means not the direct author but that of a work translated to
      make the book in question. Suddenly, the "author" tag, which
      used to allow you to conclude that the author of a finnish
      document must speak finnish, now can be invalidated by an
      element somewhere else on the record.
    </p>
    <p>
      Another symptom of a specification where the actual semantics
      may not be as obvious as as first sight is ordering. When we
      hear that the order of a set of records is important, but the
      records seem to be defined independently, how can that be?
      Independent assertions are always valid taken individually or
      in any order. In a server configuration file, for example, he
      statement which looks like "any member has access to the
      page" might really mean "any member has access to the page
      unless there is no other rule in this file which has matched
      the page". That isn't what the spec said, but it did mention
      that the rules were processed in order until one applied.
      Represented logically, in fact there is a large nested
      conditional. There is implicit ordering when mail headers
      say, "this message is encrypted", "this message is
      compressed", "this message is ASCII encoded", "this message
      is in HTML". In fact the message is an ASCII encoded version
      of an encrypted version of a compressed version of a message
      in HTML. In email headers the logic of this has to be written
      into the fine print of the specification.
    </p>
    <h2>
      Order in documents
    </h2>
    <p>
      There is something fundamentally different between giving a
      machine a knowledge tree, and giving a person a document. A
      document for a person is generally serialized so that, when
      read serially by a human being, the result will be to build
      up a graph of associations in that person's head. The order
      is important.
    </p>
    <p>
      For a graph of knowledge, order is not important, so long as
      the nodes in common between different statements are
      identified consistently. (There are concepts of ordered lists
      which are important although in RDF they break down at the
      fine level of detail to an unordered set of statements like
      "The first element of L is x", the "third element of L is z",
      etc so order disappears at the lowest level.). In
      machine-readable documents a list of ostensibly independent
      statements where order is important often turn out to be
      statements which are by no means independent.
    </p>
    <p>
      Some people have been reluctant to consider using an RDF tree
      because they do not wish to give up the order, but my
      assumption is that this is from constraints on processing
      human readable documents. These documents are typically not
      ripe for RDF conversion anyway.
    </p>
    <p>
      Conclusion:
    </p>
    <p>
      Sometimes it seems there is a set of people for whom the
      semantic web is the only graph which they would consider, and
      another for whom the document tree (or graph if you include
      links) is all they would consider. But it is important to
      recognise the difference.
    </p>
    <hr />
    <p>
      In this series:
    </p>
    <ul>
      <li>
        <a href="RDFnot.html"><i>What the Semantic Web is
        not</i></a> - answering some FAQs of the unconvinced.
      </li>
      <li>
        <a href="Evolution.html">Evolvability</a>: properties of
        the language for evolution of the technology
      </li>
      <li>
        <a href="Architecture.html">Web Architecture from 50,000
        feet</a>
      </li>
    </ul>
    <h2>
      Not put in yet:
    </h2>
    <p>
      <i>.@@@ RDF does not have to be serialized in XML but ...</i>
    </p>
    <hr />
    <p>
      <a href="Overview.html">Up to Design Issues</a>
    </p>
  </body>
</html>