-----Original Message-----
From: Eric Armstrong [mailto:eric.armstrong@eng.sun.com]
In conversation with Eugene, he's been asking
the question: Are multiple parents for a node
really necessary? Basically, if that concept
adds complexity, maybe we should plan on not
implementing it.
I have felt that it was vital, but so far have
not presented a compelling enough use case to
settle the argument. This message is an attempt
to do so. (And it led to another interesting
insight that prompted me to jot down my thoughts.)
<BIG SNIP>
Multiple parents for nodes translates to clones or references – the same
node can exist in multiple places in the “hierarchy”.
Unix
=====
Unix supports links to files and to directories, allowing the same file to
exist in multiple places. Put it in the project that spawned it, into
folders for tools that use it, into a current work folder, or move or add
links as a file moves through a workflow process.
Windows
=======
I put a folder in my start menu for web tools. In that folder I place
shortcuts (the closest Windows comes to links) to all of the various tools I
use when working on HTML and web support. Then I place a link to the folder
on my desktop to provide rapid access to a set of tools which are installed
in yet another place in the directory structure.
Literate programming editor (LEO)
========================
Edward K. Ream has written an outlining editor for doing literate
programming (LEO ). The editor is patterned on an outliner called MORE done
by Dave Winer who later did Frontier. Ed and Dave are convinced that an
outlined requires clones.
In use, as Ed begins to think about a feature, he builds a description of
the feature in a features section. As more aspects of the feature emerge,
they are added to the feature list.
When it comes time to plan a release, features are reviewed and some are
selected for partial implementation. This uses clones of the appropriate
nodes from the feature description.
As the time arrives to work on the implementation of a certain feature, it
gets cloned into the ToDo list section. As code sections are identified,
nodes in the ToDo list section describe the changes needed, and the code
sections are cloned into the list.
Now he can work from the ToDo list making changes to the cloned code nodes,
which changes the single node in its place in the code. Test entries are
created in the ToDo section as the changes are made.
When the tests are complete, the nod form the ToDo section is moved to the
change history section for the upcoming release.
If defects are discovered later, the node can be cloned back into the ToDo
list and all changes reviewed.
Clones are essential and very valuable.
MaxThink
=======
MaxThink is a DOS outliner and one of the best IMO. It supports clones.
Ecco
====
Ecco is an outline based PIM with folders as an organizing concept. An entry
is owned by a specific folder but may be cloned to others as needed. When an
entry is cloned, the ancestors are shown in a different color and cannot be
edited in the target folder. This makes it possible to plan a project in one
place, put pieces of it into folders for the people who have to work on it,
and place [Garold L. Johnson] another clone in the type of activity (phone
calls, letters, etc.) and yet another clone in the schedule. Add priority
folders and it now becomes possible to organize your activities in as many
ways as is convenient. You can look at the project, the people, the task
type, the priority, the status – whatever works.
How humans organize
===============
We organize information into as many different structures as necessary,
focusing on the structure that works best at the time, depending on our use
for it. Every real thing fits into as many taxonomies as its attributes
support. We organize in multiple hierarchies. This is one of the reasons
that some people consider multiple inheritance to be essential to object
oriented programming.
One of the common filing systems recommended uses a primary hierarchy for
physical filing while placing indexes at all other paths that are needed. If
you start looking for something, follow a path and don’t find it, you go
back and add references to the path you tried. Over time, everything that
gets used at all frequently get all of the common access paths defined so
that the item gets found from any reasonable starting point.
I conclude that multiple parents, clones, references or whatever we call
them are an essential part (as opposed to an accidental part) of knowledge
management, and are thus a requirement in any serious system.
Thanks,
Garold (Gary) L. Johnson
DYNAMIC Alternatives <http://www.dynalt.com/>
dynalt@dynalt.com <mailto:dynalt@dynalt.com>
This archive was generated by hypermail 2b29 : Wed Jan 03 2001 - 03:01:25 PST