I would like to suggest that, for HTML-3.0, anything that can be a
destination anchor can have KEYWORDS associated with it. This will
be in addition to the META element which is used to associate keywords
with the whole (or the top of a) document. Here is an example:
(in doc1.html)
<P ID="potmac"
KEYWRODS="Boston Harbor, Potomac, River"
TITLE="Potomac River">
(in doc2.html)
<A ID="missi-river"
KEYWRODS="Mississippi, River"
TITLE="Mississippi River">
What I would like to do with these keywords is to put a collection of pages
together (let's call it a book) and build an index for this book and store
the index in, say, http://myhost/mybooks/book1/book1.bookindex.
Book1.bookindex looks something like (suggestions welcome):
1 doc1.html#potmac Potomac River
2 doc2.html#missi-river Mississippi River
Boston Harbor, 1
Mississippi, 2
Potomac, 1
River, 1 2
The first half maps an anchor to a title (also gives a numeric ID to
be used later). The second half associates every keyword with a list of
anchors (since a keyword can not have commas, a comma is used to indicate
the end of the keyword).
One can then build a browser that does ``user-friendly'' search on book1
using a interface that looks like the Search command in Windows Help
which looks something like:
+=================================================================+
| Book index for http://myhost/mybooks/book1/ |
+=================================================================+
| Riv.. <-- User types here |
+---------------------------------------------------------------+-+
| |^|
| Boston Harbor | |
| Mississippi | |
| Potomac | |
| > River <-- Highlighted elem moves as user types | |
| |v|
+---------------------------------------------------------------+-+
| |^|
| Mississippi River <-- List changes as highlight above moves | |
| Potomac River | |
| |v|
+---------------------------------------------------------------+-+
The middle pane is a scrollable list so the user doesn't really have to
come close to spelling the desired keyword!
This approach is not attempting to solve the big search problem but to
facilitate searches when only a relatively small number of pages are
involved (like a book or a manual, imagine if you can do something like
this to search for the word CLASS without getting too many hits).
The bookindex files, when compressed, may be small enough to ship around
easily so you can have a html page linking book titles to bookindex URLs.
Also, the bookindex can be generated quite easily. The browser is not
hard to write either.
--
Bill Cheng // UCLA Computer Science Department // (310) 645-8328
4667 Boelter Hall // Los Angeles, California 90024 // USA
william@CS.UCLA.EDU ...!{uunet|ucbvax}!cs.ucla.edu!william
WWW Home Page: <URL:http://bourbon.cs.ucla.edu:8001/william/>