One of the enduring joys of having a pulpit, such as is provided by the XML-Deviant column, is the chance to ride one's own hobby horses. While the point of the XML-Deviant column is to represent the most interesting and important facets of debate within the XML development community, it also affords me a chance to promote or demote things, based upon my sense of what is true, good, right, or useful.
RDF is one of my hobby horses, a technology which I have consciously sought to promote in these pages, fairly and honestly, whenever the occasion presented itself. I have written often about RDF ("Go Tell It On the Mountain", "RDF, What's It Good For?, "RPV: Triples Made Plain", and "Creative Comments: On the Uses and Abuses of Markup"), as well as about technologies built on top it ("If Ontology, Then Knowledge: Catching Up With WebOnt" and "The True Meaning of Service"). My goal has generally been to suggest that RDF is a good choice for particular kinds of data representation tasks; that it is a tool every web programmer should know how to use to her advantage. The particular kinds of data representation problem to which RDF is suited include those in which some or all of the first-class objects are web resources (that is, things which have a URL), or those in which some or all of the first-class objects are amenable to being named with URIs. That is not an inconsequential problem domain.
RDF is frequently described as a tool for representing knowledge, but therein lies a hornet's nest of deeply conceptual questions, better left undisturbed if you can help it. Sometimes, however, the hornet's nest has to be overturned because it is necessary to address the deep conceptual questions directly. The formal blessing of RDF specifications, especially "RDF Concepts and Abstract Syntax", is a case in point. In what remains of this column, I review some of the vigorous debate surrounding one troublesome section of this document, section 4.
The Social Meaning Debate
As it so happens, a meeting will be convened in Boston tomorrow, 6 March, 2003, the day after I am writing this column, to discuss the issues raised by section 4. The agenda of this meeting is publicly accessible, so if you're very interested, you can follow along in detail from afar. I present the high points of that agenda in the next few paragraphs.
The RDF Concepts and Abstract Syntax specification (hereafter, "Concepts") says that an RDF graph ("a set of triples", according to Concepts 3.1) has two kinds of meaning: a formal and a social meaning. The formal meaning of an RDF graph results from applying the RDF semantics, based on model-theoretic semantics, to the graph. But, as both Concepts and the RDF semantics documents suggest, there is another aspect to the meaning of an RDF graph, the social meaning. The social meaning is the place, one might say colloquially, where RDF meets and interacts with the real world. The debate about the social meaning of RDF, and the subject matter occupying much of the agenda of tomorrow's meeting, concerns whether and, if so, how the social meaning of RDF should be formally specified in the Concepts document.
In order to provide you with a better grasp of the issues involved, I want to quote the most relevant parts of section 4:
When an RDF graph is asserted in the Web, its publisher is saying something about their view of the world. Such an assertion should be understood to carry the same social import and responsibilities as an assertion in any other format. A combination of social (e.g. legal) and technical machinery (protocols, file formats, publication frameworks) provide the contexts that fix the intended meanings of the vocabulary of some piece of RDF, and which distinguish assertions from other uses (e.g. citations, denials or illustrations)…
The social machinery includes the form of publication: publishing some unqualified statements on one's World Wide Web home page would generally be taken as an assertion of those statements. But publishing the same statements with a qualification, such as "here are some common myths", or as part of a rebuttal, would likely not be construed as an assertion of the truth of those statements. Similar considerations apply to the publication of assertions expressed in RDF.
An RDF graph may contain "defining information" that is opaque to logical reasoners. This information may be used by human interpreters of RDF information, or programmers writing software to perform specialized forms of deduction in the Semantic Web.
The debate about these issues is very complex, primarily because it crosses, and then recrosses, many distinct but related disciplinary boundaries. It has recently crystallized around an extended set of comments and reactions to section 4 by Bijan Parsia. Parsia's general objection to section 4 runs something like this: given that formally specifying "social meaning" is an incredibly complex task (and not one especially relevant to the Working Group's charter), made only more complex by trying to specify it in relation to a formal system such as RDF, Concepts section 4 is, to quote Parsia, "vacuous ... it doesn't really specify anything and thus can be ignored ... or it's dangerously underthought and underspecified".
Parsia also makes the relevance argument, focusing on section 4.4 of Concepts, which says that if you publish RDF content you thereby "commit [yourself] to the mechanically-inferred social obligations" of that RDF. Setting aside the ambiguity of "mechanically-inferred", surely, as Parsia notes, that claim falters: either because the W3C lacks the grounds upon which to make such a claim authoritatively, or because it is intended to be merely informative, in which case it is, as Parsia says, "probably false".
The Unresolved Issues
The specific issues over which some consensus must be formed before Concepts can become a formal W3C Recommendation include the following:
- how or whether the meaning of URIs (when used, for example, as RDF predicates) is defined and defined authoritatively -- and whether, or to what degree, Concepts must make this explicit (see Concepts 4.3 Authoritative Definition of Terms);
- what Concepts means by saying that "an RDF graph may contain 'defining information' that is opaque to logical reasoners", that is, what is the meaning of this "defining information"? (see section 4.2 Social Meaning);
- how one asserts, or refrains from asserting, RDF statements (see Concepts 4.1 Asserted and Non-asserted Forms);
- how one specifies the meaning of an RDF graph, which presupposes some position on the relation of an RDF graph's formal meaning, social meaning, and social meaning of its "formal entailments" (see Concepts section 4.4 Interaction Between Social and Formal Meaning);
- whether non-RDF contexts, including various technical details of the Web's architecture (protocols, file formats, and so on), are sufficient to "fix the intended meanings of the vocabulary of some piece of RDF" (in the words of the Social Meaning Discussion Agenda), and, more crucially, whether the social meaning of RDF assertions is a function of the intention of the speaker ("speaker meaning") or is a function of the meaning of the assertion itself ("sentence meaning") (see Concepts section 4.2 Social Meaning);
- whether "publishing" RDF is sufficient to obligate one to its formal or social meanings and, further, what acts or ommissions constitute a party as the "publisher" or as the "asserter" of RDF;
- finally, what relation there is between RDF's meaning, whether social or formal, and legal contexts and considerations (see Concepts section 4.5).
In short, there remain significant issues to be resolved before RDF Concepts progresses in the W3C recommendation process. There are at least two different kinds of difficulty. First, these issues (knowledge representation, the relation of formal and social systems, the meaning of meaning, and so on) are simply very difficult. If they were easy, we would have had something like RDF a long time ago, or at the least there would be even more unemployed philosophers and logicians around. Second, these issues cut across areas internal to the institutional politics of the W3C and its constitutents. RDF is in a very real sense, as my grandmother would have put it, the red-headed stepchild of the W3C. Without the active support of Tim Berners-Lee, RDF might have suffered a far worse fate than controversy and debate, namely, obscurity and abandon.
Further, these issues are complicated by the fact that Berners-Lee both participates actively in the development of RDF recommendations, in which he is one among equals, and is also able to speak, ex cathedra, as the W3C's Director. Berners-Lee has participated in the social meaning debates so far, and his position -- which he has so far expressed outside his institutional role and authority as W3C Director -- is worth mentioning here. In his view, Concepts requires a statement of "what an RDF document means" and, further, that statement is an easy one to make.
Also in XML-Deviant
This seems to fly in the face of the debate about RDF's social meaning. In other words, if it's easy, why does it seem so hard? Why are so many smart people having such a hard time doing an easy thing? One option, after all, is simply to refrain from taking a position about these very difficult issues. But that does not appear to be a viable option at this point, given Berners-Lee's view that such a statement is required. Further, we must distinguish carefully what an RDF document means from "explain[ing] how to use it", which is not required of the Concepts specification, in Berners-Lee's view.
In my view the only thing which is required is a clear statement of how the formal meaning of an RDF document is to be arrived at. While it is true that, in various social contexts, this formal meaning may well have a further social meaning, specifying that social meaning is a very difficult thing to do well. It is also outside the scope of RDF recommendations per se, and may well be pointless insofar as social meaning is typically not amenable to authoritative pronouncement, even from institutions which are legitimately entrusted with the relevant kind of authority. Even the U.S. Supreme Court, which has the authority to determine a kind of social meaning (the legal kind), did not arrive at that position solely by the fiat of another U.S. institution, but, rather, by a historical, evolutionary process. Whatever social import RDF as a whole, or any specific RDF statement, has will be something which is determined not by the fiat of any one institution, but by the same organic, historical social process out of which social meaning is always formed.
Social meaning is not just a problem of RDF
2003-03-10 07:40:33 David vun Kannon
XBRL (eXtensible Business Reporting Language) also faces the problem of social meaning directly. The taxonomies of XBRL contain explicit support for grounding XBRL concepts in opaque (to the computer) references and human readable labels.
e-Commerce, process control, security, etc. are all areas where data flows directly out of the real world into the Semantic Web. These sites are the sensoria of that web. I believe that agreement on social meaning of web documents will flow outward from these locations to more and more abstract kinds of documents, over time.