Hi, 2010/3/1 Jesse <pianohacker@gmail.com>:
Besides, this could be a long-awaiting chance to finally get rid of Zebra. The obfuscated code and quirky architecture make adding features and troubleshooting problems incredibly difficult, among other problems. If Solr is genuinely better, why keep Zebra?
To avoid *requiring* the full Tomcat & Java stack with every Koha installation, along with its obfuscated code and quirky configuration system. :) This isn't meant as a serious knock against Solr. I'm just pointing out that there would be tradeoffs if we were to just abandon Zebra, including losing a slim, fast search engine capable of indexing arbitrary XML that speaks Z39.50 and SRU/W natively with fewer moving parts than Solr. In a perfect world, I agree that it would be best to settle on one search engine and eliminate an installation option, but I don't think that we'll ever get there; there will always be another external engine that is newer/faster/shinier/better-supported than whatever Koha uses at the time. It's also important to identify the source of the problem - in the case of phrase searching, Zebra is perfectly capable of doing it, but there are a number of flaws in C4::Search's notion of query parsing that stand in the way and that would still apply if we ripped out Zebra and replaced it with Solr. Refactoring C4::Search in 3.4 will get us most of the way, and doing so with the idea of supporting Solr would keep us honest as we redesign the API. However, I would be quite content if the end result for 3.4 was simply a cleaner front end to Zebra. Regards, Galen -- Galen Charlton gmcharlt@gmail.com