Development Videoconference 20 Feb 2015

Attending: @jeromekelleher, @dannycolligan, @mollyzhang, @Naburimannu

Discussion of open PRs:

• versioning framework (https://github.com/ga4gh/server/pull/163): defer until need shown
• backend exceptions: good approach, looking for simplifications
• multiple test buckets: perhaps slightly premature, but good. Side discussions of tests vs benchmarking, defining buckets, pulling in Google GA4GH examples as integration tests.
• ga2vcf / ga2sam infrastructure and ProtocolFactory: our end goal is to take in ga4gh server command line arguments and get out VCF or SAM files. This is (1) simple, useful, and verifiable, and (2) gives us a way to deal with reads and a path towards SAM integration. VCF is a longer-term target, and pysam has some issues at the moment - but we need BAM / indices / HTSLIB from it.
• call set search: good approach, want more explicit end-to-end testing and minor test cleanup; possible issues with uniqueness guarantee of callSetId - need clarity from schema group.
• data object collection: we have problems getting more data types into our server, want a generalized approach to representing all of this rather than having to implement similar boilerplate for each type.

backend.py's Generators need to do similar iterations, and callSetSearch has more-or-less reimplemented that iteration, so we're approaching a point where we're ready to generalize. This area has been error-prone. In order to properly test references and read sets, we need to ingest that data, which is tied in with the generator structure as well.

Given a temporary lull in active contributors we're mildly relaxing review rules: one +1, delay 24 hours if you suspect things could be controversial, continue to require merge from a different organization than the author.