I want to be able to easily review and highlight sections of
documents.  I'm wondering what tool would be most appropriate. 

Of the alternatives that I can think of, I suppose that a database on
my local machine that made it easy to cut and paste key word in
context information automatically or to manually highlight and cut and
paste document text together with a link to the document and page
would work.

Another alternative would be to use a clustering engine (vivisimo) to
pull the relevant information based on keywords and a faceted
thesaurus and assemble a summary text based on clusters would be
useful.  (But I'm not unsure how useful--If the answer is that
vivisimo completely answers this need--then my search request would be
completely fulfilled).


Another alt-- A simple and useful solution for linux might integrate a
html to text convertor, a site download tool, and a revision control
system based on differences.  It might paste the page address into the
document to track the original reference text and use grep to search
the database on the local machine.  

I currently use netscape and simply mark the originals while printing
the pages and highlighting them in ink--perhaps some
simple techniques and additions might allow for an extended capability
to highlight and/or bookmark paragraphs within a document as well as
complete documents.  

I'm looking for answers that help me to explore
and capture solutions to such issues.

I wonder if all you need is a html editor like Amaya.

Then add a link in the document to your highlighed area's.

Save those links to highlighted area's for future ref.

Adding a link using an html editor is a good start.  
It addresses the first issue--linking to pieces of information.
Building on that idea, a folding editor, fe, would allow for large
scale editing of flat documents by allowing them to be collapsed and
sections could be easily expanded with the convenience of point and
click.
The html editor idea also spawns another thread--the collecting a set
of search results on dogs say, collecting the results, and building a
concordance to the corpus of information.
Its those revelant bits and pieces of information that we'd like most to see, eh?