Hi, We read that "its benefit to cost ratio is very low" [1]. In our experience there is very little cost, so would the benefit be even lower? Running countless of link analysis iterations takes many hours but running the loops job with a depth of two takes much less time.
It may be computationally expensive but the iterations of link analysis (plus writing back to the whole CrawlDB) consume a _lot_ more I/O time. Can anyone (Dennis?) provide some more details and explain why it's discouraged in production systems with billions of link nodes? [1]: http://wiki.apache.org/nutch/NewScoring#Loops Thanks
