About the model: Question 1
1. One article I read about the semantic web defined “resource” as “what you get when you click on a URL.” My use of the Web leads me to assume that when you click on a URL you always get only one thing; in other words, that a URL is a one-to-one link. What we need for our data are one-to-many links. For example, you need a URL that stands for a work, and when you click on it, you need to see all the available expressions of that work (e.g. the translations into various languages, the annotated editions, the illustrated editions, etc.). When you click on an expression, you need to see all the available manifestations of that expression (e.g. the online version, the text version, etc.) My friend Sara has argued that many one-to-one links can have the appearance of a one-to-many link (did I get that right, Sara?) Do you agree? Should I stop worrying about this? Should I be careful to use a particular kind of representation for a one-to-many link in RDF?
“Serialisation of binary associations is simple, at least in English and similar languages, but automatic serialisation of associations with three or more role players becomes difficult if not impossible. That is another reason why binary associations are so widely used”–The Topic Maps Handbook by H. Holger Rath, on the web at http://www.empolis.com“Well, I think we need to define that ‘one’ thing. You can well define it as one list of things, which usually is what happens, or it could be one relationship between many, and so on. The URL is there as an addressing mechanism for this to happen, and as you’ve pointed out it’s one of those big issues in the RDF world. In the Topic Maps world we don’t have this problem because we don’t use URL as *the* semantic of the relationships we build. For example, if RDF points to ‘http://some.example/’ as RDF:about, do they mean the RDF semantics are about *that* website, or does that website *represent* some other notion (like a topic), or is the website or the organization it belongs to somehow mixed up in the aboutness? These are not huge unsolvable problems, but enough to cause ambiguity and confusion, which of course is fatal in computer modelsA URL in itself can *mean* many things, and many-to-many, one-to-many or many-to-one are all fine examples of a URL. Let’s try a few examples ;
one : Monteverdi, composer – http://example.com/composer/monteverdi
one-to-many : Monteverdi’s 1610 vespers – http://example.com/work/123456789
one-to-one : A particular recording of the 1610 vespers – http://example.com/record/2345
many-to-one : recordings of “Dixit Dominus” –http://example.com/work/123456789/2/recordings
many-to-many : 1610 vesper psalms recorded : http://example.com/work/123456789/type:p/recordings
Some might argue that these URL aren’t unambiguous, which is true ; they provide some semantics to exemplify that their use. These URLs can equally all be expressed as http://example.com/%5Bsome huge number here]. All of them.The point here is that a URL can represent anything you want. There are no restrictions on them, no one-to-one, many-to-many or otherwise. The only limitation on URLs are their resolvability at the hosting systems. If the server you’re pointing to in http://example.com/work/123456789/2/recordings have no idea what that means, then that might be a problem. The funny part here is that in RDF there is no requirement for URLs to resolve to anything ; you can create fake URLs all the time, as much as you like. RDF state that resolvability is a good thing, though.In the Topic Maps world we use what’s known as Public Subject Indicators (PSI), which are URLs that are 1) publicly published with 2) a guarantee of maintained persistence, and 3) indicators of a subject (meaning; to get to the semantics of that URL, resolve the URL to read what it means)”–September 11, 2007 email from Alexander Johannesen.