Search is getting a speed boost in Version 0.40. See the page Discuss Version 0.40 DevTrack for more information.
-- AaronGraves Mon Mar 10 19:15:08 UTC 2014 (69.4.28.21)
Also in Version 0.40, any search request will be denied unless the HTTP referer is set to match the domain the script resides on. Additionally, if the HTTP referer is a blank search command with no additional data, the request is denied.
Though it's not a huge deal, I've noticed right here on the main Aneuch wiki site that some automated scripts were hitting the search command with a bunch of random "spammy" type phrases. This could potentially lead to a denial of service condition. Since I've set the script to deny those requests with a 503 error, this activity has completely stopped.
-- AaronGraves Tue Apr 1 20:44:33 UTC 2014 (69.4.28.21)
For search, I need to rethink how the results are weighed. Currently the most matches go to the front of the line. While this is all well and good, just because a particular page has a large number of occurrences of the key word(s) you're looking for doesn't mean it'll be the most relevant.
A great example is if you search Aneuch's own wiki site for the term "install", the page Installing is at the very bottom of the results (and at present, that search returns 18 pages in its results list). So for someone looking for help on installing Aneuch, they aren't exactly going to get much help through search (unless they look through all 18 results, but most people will probably give up after the first two or three).
I want to refactor search to apply some sort of relevancy test. Something so that in the above example, a simple search on the term "install", will produce the top result as the likeliest most relevant page.
Perhaps something can be gleaned from this: Building a Vector Space Search Engine in Perl
-- AaronGraves Tue Jul 8 16:34:43 UTC 2014 (69.4.28.21)