Thinking about searching with Google’s Custom Search Engine today led me to realize that it would be very helpful if search engines like this allowed more flexibility in how results were presented.
In the case I was considering today, it would be cool if I could do a regular spidery harvest of content, index it as usual, but then present the results in a different way. An example: items harvested from the MDL-SSR (such as this) should be able to point back at a different page (such as this) and include a link to a related graphic (such as this). This could be accomplished either by leaving “meta” hints behind in the HTML header of the to document retrieved or by writing a transform procedure to munge one set of URLs harvested into a different form for display and linking.
I wonder if anyone has done this. This is quite different from the Public Knowledge Project (PKP) approach of harvesting only structured metadata. It could maybe be built into the Heretrix project or a related indexing tool.