And as you can see from the URL Sam posted, its actually working! The search engine is indeed doing exactly what it should. I'm adding each blog document (blosxom-style text files) with the category as the path category that blosxom uses, and also the date format "path" (/2002/12/15) and so it instantly finds documents based on these paths - no filesystem access whatsoever (except Lucene's API hitting its index files). Super fast.
Querying can be by full document, title, category, date, or permalink. And the query works from the "path" (category or date) downward making it quite powerful. Sam already has this going on his site to some extent.... my extension is adding the title field and storing the blogs within Lucene. This is all automated with Ant using my <index> task.
Note: the app is currently under development but its working nicely so far.