Notification thread


(Feel free to add yourself here following the meeting - the list is still missing many names)

Discussion Topics

  • "Wishlist for Solr Roadmap" (Ishan, 5m)
  • "Solr ML Presentation" (Dave Hastings)
  • Discussion about SolrCloudClient setDefaultCollection(), discuss options. (Eric)
  • Overseer: not using it for ConfigSet operations – SOLR-16542 - Getting issue details... STATUS  (David)
  • "Solr+Zeppelin Demo" (Jason)

Post-Meeting Summary

We started the hour with a short demo from Dave Hastings on how he used Solr's MoreLikeThis capability in a novel way to identify and surface connections between researchers whose work was substantially similar.  The work was really well received, and several folks encouraged Dave to develop it for a conference talk so they could hear about it in more detail.

From there we spent a few minutes trying to unblock Eric Pugh's work on SOLR-10466.  The group wasn't too opinionated in terms of general approach, except for the general suggestion that it'll likely be easier to make a decision after trying any of the approaches on a sampling of test files to get a sense for its pros and cons.  Eric resolved to do that and resume discussion once that's done.

Following Eric, Ishan presented a short slide deck on his "wishlist" for Solr community building.  Discussion of this and related topics took up most of the hour.  Rahul Goswami joined in towards the end to +1 the focus on community building and new developer experience.  He went on to share his perspective on contributing: that he's started several times to put together code contributions, but felt pressure to meet a very high bar before sharing anything (a clean working patch, with tests and documentation).  This triggered some discussion of "Yonik's Law of Patches", and how we might center that more in our dev-docs geared towards new contributors.  The lengthy contribution guidelines we publish seem to scare people away; maybe it should be simplified or something.

We discussed using "Crave" for doing builds on a remote beefy machine.  It's integrated into PRs now.