Forum OpenACS Development: Response to Oracle SE vs. EE

Posted by David Eison on
Janine is right, Intermedia's "rebuild online" isn't part of SE on 8.1.6.  Haven't been able to find a comparison for 8.1.7 but no reason to believe it's changed.  Heard a rumor that Intermedia indexes would automatically be kept up to date in 8.1.7, but can't find any documentation to support that and don't have 8.1.7 to play with.  Phong or M.Bryzek would probably know.

Materialized views were also being investigated last I checked, and partitioning looks to me like an EE feature that is worth exploring.

Does silence when asked about bitmapped indexes indicate a lack of testing?  Nope, it indicates that the person who knew the answer to the question didn't read that message, and that there wasn't anyone assigned to go get answers from other people at aD if they didn't personally know the answer.  There really are simple logical explanations for this sort of thing: Question didn't get answered != no answer exists.  I *believe* the answer is that the bitmapped indexes were added due to lab performance testing, and that they aren't critical to performance but do help a bit, so they should be kept around if you happen to have EE (and, since aD has EE, that's what is used).  That's third-hand knowledge, though, which is why I have the "i believe" weasel words up there.

As for better Intermedia indexing, it's not hard to modify im_convert to do stemming or fuzzy matches and stuff.  I did it for a client at aD.  It slows down the queries by a noticeable amount, but it's probably worth it unless search traffic is very high.  I think a lot of the problem with site wide search is that it's hard to tell how good or bad a match is.  Unfortunately this is a feature of Intermedia - it's hard to interpret the numbers it returns except as a relative weighting within the current search, so one could assume the first hit was 100% and normalize to that, but one can't tell how good the first hit is (i.e. "Sorry user, there were no great matches, try modifying your search" vs. "Hey user, here are 100 links that may have something to do with your search, and this one at the top is the best match, but I won't tell you how good a match it is").

If somebody is feeling motivated, 2 things that would be good are:
1) trying to figure out how to answer "where do these results stop being relevant?"
2) modifying the search results page to show relevant excerpts from the page rather than page summaries.

Those two things were on my back burner a few months ago, but they never made it off the back burner, unfortunately.