The idea that a link represents “friendship” is so bizarre it had to come from some geek’s stunted view of social relations.
A link is an association, the literary sense of that word, but only the context of the link can provide the meaning, the implication, the “spin.” To engineer friendliness into a link, even on a blogroll, demonstrates a profoundly impoverished social imagination. How come so much web psychology seems to have been theorized by seventeen year old boys who don’t get out enough?
Obvious disdain for the premise aside, Joseph has reached what is the heart, the most key element, of the Semantic Web — how do we capture the context of information, because it is the context, not the data itself, that brings in semantics.
Lately, with FOAF and other uses of RDF, there is an assumption that if we just capture enough metadata, identified uniquely by URI (Uniform Resource Identifier) and throw it together, documented with RDF/XML, we’ll have the Semantic Web. More, if we just create enough web services that use this data, we’ll have the Semantic Web. The Semantic Web is about technology.
This couldn’t be further from the truth — a link is a link is a link. The Semantic Web isn’t about technology, it’s about people and about communication.
Saying that the link is a unique representation of a person, and that the existence of this link within a FOAF file denotes that this person is a ‘friend’ of the person who put it there, without understanding the context behind this link being within this file or representing this person, all we have is a pairing between what could be a useful link and an ambiguous and somewhat overused term.
As serendipity, my old friend would have it the release of Kendall Clark’s new article at xml.com titled “Social Meaning and the Cult of Tim” covers a debate that has much of this issue at its core. The debate is between Pat Hayes, the man behind the semantics documents for RDF and Tim Berners-Lee, the head of the W3C and the founder of today’s Web.
Ostensibly, the debate is about URIs, but looking closely, reading the words, it is nothing less than the issue of this ‘context’ mentioned earlier, the semantics, if you will behind the Semantic Web.
(In addition, in his article, Clark not only introduces the topic of this conversation, he also introduces the concept that there is an unspoken rule, that one does not publicly criticize TimBL because it is on TimBL’s reputation that the Semantic Web will be built. I am unaware that such a taboo existed and if does, it’s ridiculous — anything of such far-reaching importance and impact as the Semantic Web cannot exist or depend on any one person. If debates are being weighed and measured based on TimBL’s agreement and disagreement, then the debaters are fla d, and should retire from the debates and stay home. Growing tomatoes or some such thing. )
TimBL has good arguments, but so does Pat, who impressed me when I was writing Practical RDF, and continues to impress me with the arguments I see him making in this debate, and in other related ones.
[The debate starts here, and continues on via a thread between TimBL and Pat (follow reply links at the bottom, note others also join in with excellent insight, but follow the TimBL/Pat Hayes thread all the way through, first). Another debate to read starts at this point, and covers much of the same issues.]
This debate rages around the concept of a unique URI identifying, or dare I say, ‘denoting’ a single resource on the web, and that a resource must have a unique URI in order for it to be part of the Semantic Web. This is a basic concept in RDF, and is one that allows us to create RDF vocabularies that can work together.
Consider this: Within RDF/XML there is a URI that represents me. It’s used in FOAF files, but could also be used in RSS files, in Creative Commons licenses, in any vocabulary that includes URIs representing a unique person. When you see this URI, you assume it is a representation for me, and since I’m unique, it’s unique.
Ultimately TimBL wants to fix this as a given, a standard, a law if you will. Eventually there will be a global set of URIs identifying unique resources on the Web and this will form the basis of the Semantic Web. If I read Tim correctly, context cannot enter into the equation because this makes the Semantic Web difficult to engineer.
Pat responds with:
…I insist that this stipulation of identifying one thing
isn’t sensible or even desireable. Well, at least, unless that word
“identify” means something different from “refer to” or “name” or
“denote” . What might indeed be true is that in many circumstances,
a URI somehow provides access to information which is sufficient to
enable someone or something to uniquely identify a particular thing
(that the representation accessed via that URI is in some sense
about), but even there the thing identified might vary between
contexts (such as when we use someones email address to refer to the
person) without harm. This kind of ambiguity resolved by context is
at the very basis of human communication: it works in human life, it
works on the Web, it will work on the semantic Web. Why do you want
to try to legislate it out of existence? You will not be able to, any
more than you will be able to stop people falling in love. All that
your ‘ideal design’ will accomplish is to make the architectural
pronouncements of the W3C more and more out of line with the way that
the Web is actually being used by real people.
TimBL responds with:
No. We are defining the semantic web NOT to work like natural language, but to work like mathematics.
Any system of mathematics has to be able to use symbols to denote things in the universe of discourse. You as a philosopher can perhaps handle a mathematics in which symbols denote whatever anyone likes at any point, but I as an engineer find it less useful.
When Pat writes, This kind of ambiguity resolved by context is at the very basis of human communication: it works in human life…, TimBL responds with, Yes, with natural language and peotry(sic) to which Pat replies, Never mind poetry, it works for all communication.
TimBL defends the concept of a global identifier system, primarily because it’s ncessary from an engineering perspective. Pat doesn’t necessarily disagree with the fact that’s necessary, but the assumption that there is a truth to this. As he writes:
We seem to be at cross purposes. Im not saying that the ‘unique
identification’ condition is an unattainable ideal: Im saying that it
doesn’t make sense, that it isn’t true, and that it could not
possibly be true. Im saying that it is *crazy*.
At first you might think that Pat is following the theoretical too much — the way of the semantician, the way of the linguist — until you start to see what exactly he’s saying. And he’s right. Pat is right.
What Pay is saying is that the concept of a URI identifying a single resource works, but it’s broken; but that’s alright, because it works, but don’t make any additional assumptions of truth based on this. In other words, there is a URI (the URL — the address you type into the browser, or the permalink) for this page, and this page can be identified by this URI. For the most part, as long as you and I agree on this, we can work together and create vocabularies and technologies that work together. But the concept is flawed because it does not take into account the contextof the URI — that thing that Joeseph pointed out in my comments. The “spin”.
It is the importance of the context that Pat defends, and it is this very context that TimBL says must not be taken into account, otherwise the system can’t be engineered. But it is the context that sets the assumptions we can make, and to use a universal set of assumptions is just as meaningless as to depend on a universal set of identifiers regardless of the context of their use.
Pat expanded on this in another thread:
BTW, the current usage of “resource” in the SW specifications is
vacuous: a SW Resource can be anything whatsoever, real or imaginary, on or off the Web, in the past or future, of any nature, with or without a URI. So to claim that all SW resources ‘contain’ a Web
resource sounds like it would also have to be vacuous or else would
be obviously false (depending on what a ‘web resource’ is, which of
course I have no idea about, this never having been defined or
I have no idea what an interface to an object could possibly be.
What kinds of interface do the following objects have: a grain of
sand, a galaxy, an imaginary detective ?
There’s the key, the understanding — what kind of global system can assimilate objects of such differing contexts as that of the micro (the sand), the macro (the galaxy), and the virtual (the imaginary detective)? No one system can, but no one system has to. What Pat is saying is that the current system works ‘good enough’. It works though it’s based on a broken premise of a global system of identifiers that can denote any one thing regardless of context. It’s okay that it works, and it’s okay that it’s broken — but don’t base laws and assumptions on a broken premise. Don’t attach meaning to the system, just use it.
This returns again to what Joseph said. How can we ‘denote’ a friend, just through a link labeled as such? It doesn’t take into account the context of the label and the system. By taking a set of links in a blogroll and creating a FOAF file and saying we ‘know’ each of the people listed in this blogroll, we’re taking the links out of context — a link in a blogroll is not the same thing as making a statement that I know this person, or this person is my friend. The only statement I’m making, taking into account ‘context’, is that I listed this person in my blogroll for some unspecified reason of which no one can truly make an assumption.
There may be an assumption that I did so because I read them, or that I like to read the person, or that I like the person. But that’s all this is, an assumption. Without the context behind my reasoning why I put links into my blogroll, how can one then extrapolate out that these links should then go into a FOAF file? Or vice versa? How can one extrapolate that the context of the FOAF file and the context of the blogroll are the same? Because the same identifiers are used in each?
Yes, the syntactic string representing a person may be the same in the FOAF file as in the blogroll, and because of this and the use of RDF, we can ‘technically’ combine and extrapolate this information — but without the context surrounding the use of the identifier in each case, you can’t make an assumption that the one ‘means’ the same as the other. In other words, you can’t extrapolate, meaningfully, from my URI appearing in a FOAF file to my URI appearing in a blogroll, because neither is ‘me’ — only me as I am represented within the context of each vocabulary.
Pat wrote (and I can’t find the exact email message):
How can one consider a link to be ‘the person’, when it is nothing more than a proxy, a representation that the Semantic Web requires because we have no other way to represent the person within the Semantic Web.
Within the Semantic Web, a URI is a proxy for, a representation of, something that can’t be represented any other way due to the liminations of the medium. And that’s okay, because it works. But, if I read Pat correctly, don’t add any additional ‘meaning’ to this representation other than the fact that it is a representation. To do so will perpetuate the broken premise.
FOAF can’t represent a friend, or a relationship directly. What it can do is provide a proxy for an association between two people, as marked by one of these people, and as a labeled friend, aquaintance, co-worker or whatever. It is not the actual relationship itself, and to see it as such, to treat it as such — to make it real because of this association in the file — removes the context of the FOAF file, which could have significant impact on the truth of the assertion.
Because I am listed as a ‘friend’ in AKMA’s FOAF file, does this make it real? You can’t assume it’s real just because it says so, in a FOAF file, with a link, representing a URI. It may be real — it is real to me because of my association and appreciation and affection I feel for both AKMA and his wife, Margaret — but the link, and the existence of the file, don’t make it real.
Moreover, you can’t extrapolate any additional meaning out of the FOAF file other than what is narrowly defined within the context of the vocabulary — FOAF shows that one person is making a statement that they know another person. Nothing more. Nothing less.
Pat isn’t being arbitrary, he’s making a critical point: we can only assume so much from a URI within context of a RDF vocabulary. To make additional assumptions is as false as to make an assumption that the context of FOAF and a blogroll overlap, and that my relationship to one person in this way, must mean that it’s the same as my relationship to this person in another context.
Dammit, I’m not saying this well, but it’s the very use of FOAF for other things outside of the context of this specific RDF vocabulary that forms the basis — in my interpretation, and I could be wrong — for Pat’s continued and persistent argument about the URI of an object being a proxy for that object, and that a URI has context. To ignore the context is to literally throw out the true semantics, leaving nothing in its place but a smarter, but still dumb, web.
Smarter web is okay, but I want a semantic web! I don’t care if the Semantic Web works for the technology if it doesn’t also work for the people.
By seeing the URI as a representation of an object that transcends context, we then erroneously make extrapolations, such as FOAF and the blogroll — harmless in this case, but not so harmless when you start bringing in issues of trust, and see FOAF as the basis of the web of trust.
What Pat is arguing about, the point he is trying to make, forms the basis of what is happening now. We only have a few RDF/XML vocabularies in wider use and already we’re seeing abuse because of making assertions based on flawed premises. This isn’t a semantics argument or a esoteric debate between philosphers — this is real stuff being implemented, a perpetuation of a premise that’s flawed.
I have more to say on this, later. I must read all the notes, think on this further. If I’ve mispresented either TimBL or Pat, my apologies to both and blame it on my interest and excitement about this debate.
Archive with comments at Wayback Machine