Subject: Re: LISP and AI
From: Erik Naggum <erik@naggum.no>
Date: 2000/05/09
Newsgroups: comp.lang.lisp
Message-ID: <3166856052016369@naggum.no>

* Espen Vestre
| Yes, I was puzzled by this too.  The only thing I could think of
| was the fact that bibliographic references may refer to a section
| or page range of a second document and not only a _point_ in it.

  Please ignore HTML and the WWW in any discussion of hypertext.  They
  are like bringing PostScript and laser printers into a discussion of
  literature.

  There is no worse implementation of hypertext concepts than the
  in-band anchors that requires changing both (there are only two in
  HTML) documents if you need a particular connection.  In fact, the
  HTML/WWW implementation of "hypertext" is so fundamentally flawed
  that it is probably a great disservice to hypertext to call it
  "hypertext" to begin with, and as witness, the inability of people
  to consider a bibliographic references that does not match the HTML
  _implementation_.

  Consider the glossary that defines hypertext links _to_ its defined
  terms.  Instead of myriad explicit links in each document in the
  document set covered by the glossary, the links may be defined as
  whatever matches a rule in those documents.  A rendering engine
  would load the glossary and apply the rules to highlight glossary
  terms, or perhaps make them more accessible by popping up a small
  window with the definition.  Instead of this being coded explicitly
  wherever a glossary term is used, it would all be arranged in the
  glossary, once.

  Consider the root document of a document set that contains rules for
  which documents (such as the glossary) whose rules should apply to
  which documents in the set.  Conventional bibliographic references
  expresses such relationships with annotations in the library records.
  Note that there is no unique text to highlight or click on in this
  case.  This is meta-information for the hypertext system.

  Consider the critique of a document that includes excerpts from it
  and does so using in-lined hypertext links instead of copies of the
  text itself.  Consider the meta-critique that contains a full list
  of all such references for the purposes of scholarly research and
  ratings.  The former would not necessarily be able to influence the
  base document when somebody reads it, such as to inform the reader
  adbout the critique, but the latter would, as well as responses and
  rejoinders in a debate where various authors both discuss the type
  of critique and rate them.

  Consider the continuous publication of a medical journal where it is
  incredibly important to link from articles in the past to new and
  updated articles in the future with crucial information about the
  role of the update.  A new article would typically contain a small
  passage that it updates, contradicts, etc, previous articles.  It is
  not uncommon for the journal librarian/editor to supply such links
  in a separate bibliographic unit, such as a side-bar.  A mechanized
  hypertext system would represent such links with meta-documents that
  _all_ the articles in the system would point to for updates to
  themselves.

  Consider the continuous application and concurrent drafting of laws
  and regulations in a society.  The entire legal world is intertwined
  in extremely interesting ways from a theoretical hypertext point of
  view.  Political discussions often take the form of contributing to
  a decision on what may be done within the framework of certain
  regulations, such as budgets, legal authorities and procedures, etc.
  Counter-arguments frequently attack the hypertextual nature of the
  argument instead of the textual contents in the shape of denying or
  rejecting an interpretation of such authority or its application.
  Court arguments frequently involve comparing prior applications and
  cases to present cases.  Yet, each document produced contains only a
  small number of the hypertextual links involved: the remainder are
  implicit or take the shape of "to"-links instead of "from"-links.

  In the world of bibliographic references, there is a lot more going
  on than just pointing to books or pages, or a footnote someplace
  that has an ISBN.  Trivializing the bibliographic reference to
  whatever HTML can represent is like arguing that PostScript cannot
  represent irony, so therefore it is only characters, like all other
  characters, or denying that laser printers can reason because they
  produce documents that contain reasoning.  (Nobody argues that laser
  printers can reason, just as nobody argues that hypertext links can
  be "discovered" from context and contents, which is a strawman
  argument frequently used against advanced hypertext theory.)

  For those who want to understand how hypertext started, I urge you
  to read Vannevar Bush's original article in the Atlantic Monthly.
  In particular, he discusses machine-aided annotations in a way that
  would make it impossible for anyone who had read that article to
  even think that hypertext links were contained _in_ documents --
  that is merely an implementation optimization applicable in a few
  circumstances and not at all generally.  Generally, we make links
  into, between, and out of read-only documents that we do not own or
  control.  HTML does not allow us to work with documents we do not
  own or control except by pointing at them as static objects.

#:Erik