home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Simtel MSDOS 1992 September
/
Simtel20_Sept92.cdr
/
msdos
/
hypertxt
/
hyper.arc
/
FILE28
< prev
next >
Wrap
Text File
|
1988-02-01
|
7KB
|
127 lines
Hypertext -- Shouldn't it be hierarchical?
==========================================
Hypertext should be hierarchical for the following two reasons:
┌───────────────────────────────────────────────────────────┐
│ Hierarchies are essential to browsing │
│ Hierarchies are essentail in communicating knowledge │
└───────────────────────────────────────────────────────────┘
These are the primary reasons for organizing information within hypertext
formats. The central idea of hypertext is organizing information in
non-linear formats, which support multitudes of interconnections. Such
formats allow users to access whatever they want from wherever they start.
However, there are a number of problems with this concept, such as:
TRAVERSAL LIMITS Sand on a beach in California has a physical hypertext
================ connection to a grain of sand on a Florida beach. But
who would want to trace the connections, or who cares?
How many nodes The problem with hypertext-sand is not the concept of
will you look interconnections in objects, but the number of nodes
at? traversed. The same is true with information hypertext.
There exists an upper limit in the number of
intermediate nodes that people will examine before
seeking alternative ways to find what they want.
What is that limit . . . three nodes, five, or maybe
Limit of 5? seven? I think the answer is less than five.
If, after five tries, the user is no closer to the
desired information than at the beginning of the
process, most people will give up the search. Who would
continue using a hypertext system that just presented
five frames of unwanted information? Outside of trivia
experts, most people would not tolerate such behavior in
a system.
PARALLEL NODES On the other hand, if each frame presented some portion
============== of the desired information, a user might browse through
10 to 15 parallel nodes before wondering why it wasn't
put altogether as an idea unit in the first place.
Consequently, although there is no reason why a
hypertext topic can't be split into separate nodes for
each paragraph, line, or word; in practice, linear
information should be left linear. Idea units seem to
be the natural boundary for each hypertext topic.
INCREMENTAL The third approach to providing increasing amounts of
HYPERTEXT? desired information with each subsequent node seems to
========== be a natural reward mechanism to induce continued
browsing by a user. That implies a hierarchical order
within the nodes browsed.
Unfortunately, the originator of each of the nodes may
not have anticipated the specific information desired or
the sequence of nodes browsed. Why?
CIRCULAR Most creators of hypertext see nodes as idea units with
HYPERTEXT cross-references, and not as hierarchical transmission
========= units. Furthermore, imbedded buttons and page-based
hypertext systems are not suited for more than three or
four branching points per display. This hardly makes
for efficient hierarchical systems.
Actually, that's the primary problem with most hypertext.
Hypertext cross-references in a body of knowledge tend to
be circular <FILE77 DEFINITION> rather than directed.
HIERARCHICAL The problem with most hypertext systems is that they
HYPERTEXT aren't hierarchical <FILE74>. Yet, hypertext needs to be
========= hierarchical to enable people to pick up from portions of
the structure the overall organizing patterns of that body
of knowledge. Hierarchies are predictable; networks
aren't.
Another way of saying this is to say that hierarchies
are prime communicators of ideas. Whereas people can
handle information in more complex forms than a
hierarchy, it is difficult to communicate anything more
complex than a hierarchy to others. <FILE75 STRUCTURE>
Critical reading requires assembling ideas, integrating
them, and figuring out the appropriate context in which
the information should be placed. However, most people
don't read critically -- because it is too much effort.
However, with information organized in strong
hierarchies, readers have a sense of where they are
going, why, and how everything fits together.
For these reasons, I use hierarchies to categorize information in ways such
that readers will accept them, using the following guidelines <FILE50 RULES>:
Average number Good hierarchies normally have five subtopics. The
of subtopics reasons are -- 2 is too didactic, 3 or 4 is too
============ symmetrical, and 6 or more is too many (unless the
classification methods of subtopics is obvious to the
reader).
Completeness The set of subtopics should appear to cover all choices
============ (i.e., appear complete) as checking for exceptions is
usually the first test readers apply to any hierarchy.
Parallelism The set of subtopics should appear to divide the subject
=========== into roughly equal categories. That's the second
thing readers look for in good hierarchies.
Sequence Finally, the order of subtopics should be natural to the
======== reader.
If you follow these guidelines, you'll know the secret of good hypertext
<FILE62 ORGANIZATION>. It's simply organizing information so users can rapidly
make choices that lead to the information they desire.
Just in passing, with their experience in classification, users of MaxThink
and Houdini <FILE26 INFORMATION> know more about the essentials of hypertext
than any other group of thinkers. How about that?
Neil Larson 1/16/88 FILE28
44 Rincon Rd., Kensington, CA 94707
Copyright MaxThink 1988 -- Call 415-428-0104 for permission to reprint