Meta-Browsing on Existing Systems

[Brian Rice]

At 01:30 PM 5/23/00 -0500, Marc Santoro wrote:
>   Herein is contained a summary of information on meta-browsing gathered
>from the Tunes site, and through conversation on IRC. Hopefully this
>information can be used to spark a useful debate. This is my
>understanding:

I'll help out by providing the list with the context for this discussion:

The article on meta-browsing is "TUNES against the WWW" at 
http://tunes.org/Interfaces/tunesvswww.html

The IRC discussion in question is at 
http://www.tunes.org/files/irc/2000.0523 starting after the 08:10am marker, 
at about line 50.

A more formal definition of meta-browsing than is given on the tunes.org 
page is that it includes accessibility to meta-information concerning 
published content. This should include everything from syntactic structure 
to functions and predicates over that content. This should also be 
higher-order, so that instantiating meta-information can be accomplished 
programmatically.

What we're seeking with this is a new set of content for the article that 
reflects all of the good input from the group, and also provides an 
effective description of how *existing* tools could provide a working 
example of this concept in little or no time.

>   The Web is currently inadequate for any kind of meta-browsing. Content
>on the Web is almost completely static, and when it is not static, it is
>highly restricted and nigh-unusable. CGI allows the browser to see only
>what the script author had the foresight to program. In order for a CGI to
>allow meta-browsing, it would have to be so extremely complex, that it
>would be very difficult to learn to use. XML and DHTML are some
>interesting ideas. However, they are designed more to enhance HTML pages
>than to allow a true meta-browsing system. A page using HTML is still a
>linear page.
>   To be truly useful, our system needs to present information in the form
>of meaningful objects that can be interacted with. These objects not only
>need to interact with the user, but they should also be able to interact
>with eachother. Obviously HTML is useless because it allows no real
>interactivity. XML is useful, but it is still very limiting. XML is still
>a markup language, so we can define structured documents, but the browser
>is still limited to what the author wrote in his document. The reader does
>not get the full benefit of the authors knowledge, nor his own ability at
>processing information.
>   The problem with the current markup languages comes down to one simple
>point. They can't be arbitrarily interacted with by the user or the
>system. With meta-browsing, the system should be able to interact with the
>document natively. The document, at the very least, could be composed of
>objects that can respond to messages. These objects can reply with data,
>or could be a working example of the concept being described. And ideally,
>the objects should be a union of the two. The user could then browse these
>objects as he desires, and these objects could interact with each other to
>present the user with related information on what the he wants to see.
>   The benefits to such a system are obvious. The user could be able to
>maintain the objects locally in such a way that he can edit and republish
>them. The objects can be written in many languages, when one considers
>that most languages have Foreign Function Interfaces. If a message-passing
>object system is used, and the message format is standardized, a FFI can
>be designed for many languages that will allow them to interact in the
>document transparently. A user can use the system he prefers to interact
>with the objects, freeing the author from the tedious task of suiting his
>information to potential readers. And the user can modify the document
>without learning the details of the languages the document was designed
>in.
>   In summary, a meta-browsing system needs to allow documents that consist
>entirely of objects. HTML can not provide this. In the meta-browsing
>system, the user can interact with the objects in any way he desires, to
>find the information he wants. These objects do not have to be written in
>any particular language, existing systems such as Squeak and CommonLisp
>(Both ACL and CMUCL at the very least) provide facilities necessary to
>link the objects together. The user does not even need to use the
>language[s] the objects were designed in to interact with them; he should
>be able to use any system that provides an interface to the objects. This
>could be a "point-and-click" browser, or it could be an environment like
>Squeak or Lisp. The most important thing is that the user has the ability
>to interact with the information provided using any method he sees fit.

Sorry, Marc, for not backing you up on this one properly. Anyway, bring on 
the comments. :)

~