I plowed through both of the articles mentioned here in the last few days. I
also read some of Mimi Yim's related articles, which are written with a fun
sense of humor. I loved the inclusion of a few frightening examples of
people's organizational structures. To try to summarize some of what these
various articles suggest:

Benefits of Hierachical Structure
.People like hierarchical structures (items within files within folders)
because they give information a sense of location and sequence.
.Hierarchies tend to work best when tailored to an individual's personal
needs.

Problems with Hierachies
.Complex hierarchies are hard to construct (takes time and thought), and are
dependent on purpose.
.The appropriateness of hierarchies and sequence of items can both matter
(year/month/day and Sunday, Monday, ...)
.They break down semantically as they grow more complex.
.Different people use different terms for similar things (Tasks, Actions,
and ToDos; Notes, Memos, Items).
.Hierarchies tend to grow messy, with too layers and folders, too many
"random" or "misc" folders, and too many items stuffed into folders that
poorly match the item.
.Difficult to quickly reorganize for an ad hoc purpose (takes lots of time)
.Require placing items into appropriate locations (takes time)
.Rigidity of the parent/child model (many items could usefully go more than
one folder)

Benefits of Tags (Categories, Keywords, Attributes, Facets)
.Tags allow searches to locate items anywhere in the database
.Tags allow searches to group items together quickly based on the tags.

Problems with Tags (Categories, Keywords, Attributes, Facets)
.By themselves tags provide little sense of location or sequence relative to
the larger pool of data. (The databases feel chaotic.)
.Tags have to be added if not inherent in the data (takes time).
.Data thrown into a non-hierarchical structure with tags omitted is like
throwing items into a giant bin. You forget what's in there, and, unless you
get lucky with search terms, you never see it again unless you go through
the bin item by item (which takes time, especially if useless items are in
the bin). So ideally, tags should be added to the item early on, such as
when the item is placed in the database (takes time).
.Hard to create good tags (issues of consistency, synonyms, etc., so takes
thought)
.Tags multiple and become chaotic.

An ideal PIM might:
.Allow both a hierarchical structure and ways of tagging individual items.
.Allow items to be assigned (linked, cloned) to multiple folders in a
hierarchy.
.Allow items found in a search to easily be assigned (linked, cloned) to a
new folder.
.Allow for the automation of certain tasks, such as folder assignments and
keyword indexing.

The two programs that I have worked with that come closest to meeting this
ideal are Zoot and Ultra Recall. Zoot allows for more automation, but is
currently limited by 16-bit size barriers and lack of rtf. Ultra Recall is
remarkably advanced for a version 1.3 program, but lacks some of the
automation features of Zoot. However, its automatic keyword indexing of data
makes for lightning fast searches (but the database files become large), and
these searches and individual items can be placed in new linked locations.
And its template system is very nice for creating items with specific
attributes.

Not surprisingly, both programs are apt to bewilder anyone who gives them a
casual try. Their very flexibility forces the user to immediately confront
the challenge of creating useful hierarchical structures and good tags. To
provide a glimpse of what is possible, both programs provide sample
databases. The array of possibilities may make some users feel that these
programs provide more power than they need (which could be true for a person
who only needs to store a modest amount of easily structured information).

I think that what these articles make clear is that information storage and
retrival inevitably becomes a complex problem when large amounts of diverse
information are stored. Great, flexible tools can help, but there is a time
cost in learning to use them well, a time cost in structing and maintaining
large, complex databases, and the challenge of the database design comes
back partially toward the user when the software allows greater flexibility.

Randall