Re: [unrev-II] TopicMaps, Ted Nelson, Virtual Files, and everything

From: Lee Iverson (
Date: Wed May 30 2001 - 16:05:49 PDT

  • Next message: Jack Park: "Re: [unrev-II] TopicMaps, Ted Nelson, Virtual Files, and everything"

    In message <>, Jack Park writes:
    > [some deleted]
    >On top of that, Ted talks about Xanalogical Media, which includes something
    >called a Virtual File. Now, we are strongly reminded of Topic Maps and
    >Steve Newcomb's lectures that everything in the way of references and links
    >should reside *above* the information resource. That, if nothing else,
    >describes Ted Nelson's Virtual File. I believe that I have read somewhere
    >that word processors do something like this when one opens a document for
    >edit: a temp file is created complete with links that, during the save
    >operation, allow merging temp with the main document. After listening to
    >Ted, heck, why not do everything that way in the first place.
    >BTW: my understanding of Douglas Engelbart's Augment system informs me that
    >Augment "files" do precisely the same thing.
    >An advantage, as advertised, is this: you now only need one copy of any
    >given item in an information resource universe. You just wire these
    >objects together in a virtual file and *presto!* you've got document, mail,
    >So, what happens if one simply begins to construct all documents with
    >virtual file systems, and, um, suppose the virtual file system were to
    >implement XTM documents. One wonders just what levels of searchability,
    >readibility, and so forth (not to mention, disk space requirement) one
    >could achieve by this subtle shift in action.

    I believe that there is a problem here. I don't believe Ted "gets"
    the shifts in understanding that have taken place in moving from SGML
    and HTML to XML. In almost all circumstances, embedded tags are now
    (reading the most-enlightened XML literature) considered to be
    semantic type identifiers. XML is now painted as a document-oriented
    data modelling language. The fact that it is not very good at this
    particular job is generally glossed over. As Eric has pointed out
    many times, the real "problem" is not embedded tags per-se, but mixed
    content, in which you don't know whether you're going to get text or a
    tag as the next item (ick), and in which you are mixing semantic tags
    with formatting tags (double ick).

    >DocBook would nolonger be comprised of imbedded tags. Rather, it would be
    >an "above the document" series of URIs into a rich information
    >space. Parsable as a DocBook document, but now, given the ability to let
    >the URIs (tumblers?) pass through a grove engine, one now has the
    >capability of universal multimedia document generation, complete with
    >reusable components.

    I'm with you entirely on this one. Another way of phrasing it would
    be to suggest that we have a soup of completely reusable components
    (nodes) which are organized using a small number of very simple
    primitives into an arbitrarily interconnected graph of nodes.
    Edges in the graph may be references to either local or remote
    objects, as long as every potential object has an easily derived
    external address (a URI). Information objects are then delivered
    outside the system by serializing views (Doug's term, the same as
    Ted's virtual file) using a variety of serialization algorithms each
    of which corresponds to some external document format (e.g. HTML, MS
    Word, GIF and JPEG images, ...). These documents only actually exist
    as serializations produced from various views of the interconnected
    node databases. Sounds like NODAL!

    >Of course, one needs a mechanism to view (and edit) such an enormous,
    >heterogeneous information space. Ted gave a great demonstration of his
    >ZigZag technology that allows individual information-bearing nodes to be
    >wired together, just as beads on a string, and wired into as many
    >dimensions of information one wishes. One node, many views. An open
    >source version of ZigZag is available at I
    >am persuaded to suspect that an application of the ZigZag idea merits
    >consideration in just about any knowledge project in which one might be
    >involved, including Doug Engelbart's own Open Hyperdocument System

    I wasn't particularly impressed with ZigZag, since it didn't seem to
    give me anything that I couldn't get with generic (and
    well-understood!) graph structures and algorithms. Ted's search for
    "revolutionary" data structures seems to be too much of a barrier for
    most programmers, let alone the users of his systems.

    >Boggles the mind to think that Ted Nelson, Doug Engelbart, and very few
    >others have seen the universe this way and it has been largely ignored. I
    >am thinking that the *above the information* paradigm rediscovered by
    >Newcomb, Biezunsky, and others needs an even closer look.

    I'm not entirely convinced that all structure wants to be "above the
    information" . I'd suggest that useful information has inherent
    structure. The real issue is the flexibility of the structural
    building blocks and the ability to reference and reuse this structured
    information with a variety of higher-level structures. In order to
    reuse a birthday, I want to maintain the fact that it is So-and-so's
    birthday, no matter what the context that nugget is being used in.

    So, with your permission, I'd say that the real failure of current
    systems is that the only level of reusable structure "above the
    information" is the document. That is just way too coarse.

    Lee Iverson SRI International 333 Ravenswood Ave., Menlo Park CA 94025 (650) 859-3307

    Community email addresses:
      Post message:
      List owner:

    Shortcut URL to this page:

    Your use of Yahoo! Groups is subject to

    This archive was generated by hypermail 2b29 : Wed May 30 2001 - 16:18:32 PDT