> Lawyers refer to a case by the book and page number, and refer to parts of a case by the page numbers. It's probably not a great system [...]
It's a fantastic system (and a wonderful demonstration of the advantages of immutable graph data structures). You can read cases from a century ago and still find every cited reference. Hyperlinking pales in comparison.
It seems to me, they solve different problems: one is about finding a document Foo, the other is about finding information in a document Foo, assuming you already have Foo.
The hyperlink equivalent would be "you can find this book at this library in this town, as of the time this was written".
I imagine that in real life lawyers don't need to worry about that sort of instruction because those books are printed and widely distributed across the country, while on the web documents tend to exist in only a single location, maintained by a single person or organization. USENET might be the closest on the internet we have gotten to that sort of distribution model.
Citations can refer to the document as a whole. E.g. 410 U.S. 113 (1973) refers to the case Roe v. Wade. It's equivalent to http://en.wikipedia.org/wiki/Roe_v._Wade. 410 U.S. 113, 153 is a pin-cite to the quote "This right of privacy, whether it be founded in the Fourteenth Amendment's concept of personal liberty and restrictions upon state action, as we feel it is, or, as the District Court determined, in the Ninth Amendment's reservation of rights to the people, is broad enough to encompass a woman's decision whether or not to terminate her pregnancy." It's similar to http://en.wikipedia.org/wiki/Roe_v._Wade#Right_to_privacy.
The difference is that a citation is a reference to an immutable object. No matter what happens to Roe v. Wade, say it is overruled, the document at 410 U.S. 113 won't change. A URL on the other hand is a reference to a location. Like with the memory address pointed-to by a C pointer, what is at that location can change.
One can imagine a system of hyperlinks that behaved differently. In this system, a URL would uniquely identify an immutable document. A new version of a document would get a new URL, and servers would be required by the protocol to preserve all previous versions of the document. This is essentially the premise of Git: every blob is stored forever and different revisions are stored as deltas such that older versions always remain accessible.
It's not really hyperlinking that fails, but the fact that web sites (the ones that don't disappear without a trace that is) don't maintain their structure over time. Many don't even attempt to provide meaningful forwards, instead just dumping you back to the front page. Microsoft and Oracle are particularly egregious offenders.
What we need is a technical mechanism for embedding referenced source documents into a document, in a way that is as easy as hyperlinking to add the references and follow them. Probably also a new fair-use provision in copyright law as well.
I think it's fair to say that hyperlinking fails. Sites don't maintaining their structure over time would be the cause of the failure, but the link still doesn't work... Semantics maybe, but eh.
Agreed on having a way to embed documents. Especially in a way that supported some kind of signing. If I could have a reputable third party (web.archive.org or anybody else) sign an embedded snippet of a page and say "Yes, this was actually posted on X website at Y time" that would be fantastic.
Hyperlinks are just interactive references. Hyperlinks don't care about the nature of the reference. It is quite possible to have a hyperlink with its reference set as a book/page number.
It's a fantastic system (and a wonderful demonstration of the advantages of immutable graph data structures). You can read cases from a century ago and still find every cited reference. Hyperlinking pales in comparison.