This might more properly be a server function than a client one.
Depending on the server, there is no guarantee how the "keywords" used
to find the document correspond to any words in the document.
I can image a server marking up the document with some appropriate
highlight emphasis tags around the matching keywords. This would have
an advantage in cases where the server does sophisticated stemming,
etc, of search keywords.
A problem is that due to the stateless nature of HTTP, the server
usually does not have the context to remember what the search keywords
were when the document is retrieved. The only way I can see around
this is to encode some of the context into the URLs returned by the
search (yuck?). For example:
http://server:80/database?word1+word2
might return a list containing:
http://server:80/database/doc3458?word1+word2
http://server:80/database/doc8754?word1+word2
http://server:80/database/doc9235?word1+word2
and in turn, retrieving the docs using the above URLs would result in a
copy with the search words appropriately highlighted.
Just an idea. Comments?
Steve Putz
Xerox PARC