Hi Ian,
Many thanks for that explanation and especially pointing out the reduce method. I'm just wondering how this affects non-Nutch searcher implementations. For example, we are using Nutch for crawling/indexing, but our own application (using Lucene's API) for searching. As far as I can see, we would like to continue using the nutch document-level scoring, but retain the ability to boost fields explicitly, so that Lucene's own calculation of norm will take the field-level boost into account. Does it make sense to provide the ability to apply field-level scoring and not always blindly apply the one calculated at document level by the scoring filter to every field? Best regards, Alan _________________________ Alan Tanaman iDNA Solutions
