Think of filter in unix, where you cat file, pipe it through egrep, the awk, then sort, uniq -c, and the sort again. Something I commonly do.
Now that there are api's out there to let people access databases via the web, like doing google searches.
Now imagine a plethora of web servers providing filtering services of various types. And a front end webserver that would have an AJAX interface that would look and work similar to Visual Small Talk, where you could stack these web-filters in a series that would look like a flow chart. And then give it an input of an online database that you would send a query.
Then once you got the web filters set up how you like, you could save the flow algorithm and put a skin over it to use as a web form on your own website.
Why? Because sometimes I find the results from google worthless and the built in filter methods clunky, and if it was something I searched for often it would be able to save the search/filtering scheme.
The downside would be it would be hard finding people to volunteer memory/cpu cycles for massive sorts and the like and transferring all the data via http with POSTS and GETS would be abysmally slow.