Attending: Filipe Bento, Chris Hallberg, Demian Katz, Benjamin Mosior, Tod Olson, Luke O'Sullivan, Ben Wiens
Several PR's pending the 2.2 release have now been merged.
As mentioned above, 2.2.1 was released on 2/26 to address (primarily) some bugs in the Bootstrap/Bootprint themes.
A formula (see this thread) has been found for determining how much memory will be consumed by the filter cache. This seems to suggest that our filter cache settings will take up huge amounts of memory, but KLN team is not seeing this in practice with the existing configuration.
Ideal setting depends on many variables – hit ratio, number of records, etc. In Solr admin, there is a “Plugin/Stats” tab which shows a filter cache drop-down where stats can be found for analysis – http://example:8080/solr/#/biblio/plugins/cache?entry=filterCache. (One strategy found “in the wild”: log for three months, use that value).
It seems that a lower default would be justified, but more data is still being gathered to support that decision. At Code4lib, VuFind representatives will touch base with Blacklight developers to see if/how they are addressing this issue.
There was some discussion of the impact of multi-core setup – does the authority core really need/desire the same settings as biblio? Probably not.
Autowarm was also discussed – it may be worth testing with different autowarm settings.
Tod has checked some code into SolrMarc – org.solrmarc.callnum package has been established containing classes for dealing with Dewey/LC call numbers (and UChicago local subclasses as examples). Dewey classes are now fully independent of other SolrMarc components (no dependency on older call number code); LC classes still need further refactoring to eliminate dependencies, but this will eventually be done.
For the browse handler, normalization enhancements are done. Remaining question: do we bundle in call number handlers, and if so, how do we get access to the call number components embedded in the SolrMarc one-jar jar? Accessing one-jar content would require a special classloader. Tod's workaround: manually extract marc-importer.jar from the SolrMarc one-jar into the VuFind browse handler environment. Not very pretty, but it works. A better solution would be nice.
We still need to think more about the packaging mechanism; Tod will continue working on improving the code while we think over the “big picture” issues.
Demian will send an email to the solrmarc-tech list to see if there's interest in a conference at Code4lib to talk about some of these issues.
Demian has completed the proposed refactoring.
Tod found this documentation page. Should we update import scripts to turn off this limit? It is causing problems in SolrMarc on newer JVMs.
This setting was likely introduced as a defensive measure to prevent exploits involving excessive entity expansions in XML documents sent over the net. It's probably safe to turn off the limit in the context of SolrMarc – we just need to see if setting to 0 actually works. Tod will follow up and find someone to test this.
Demian will open a JIRA ticket and do more research; we will discuss again on the next call.
Deferred to next call. Demian found no objection to a new logo on the Villanova side, but Ben's proposed logo is still under development; he suggested that Chris and Ben touch base on some details.
JIRA has been upgraded due to a security vulnerability.
The next call will be Tuesday, March 18, 2014 at 10am Eastern Daylight Time (14:00 GMT).