Kepler 2.0 Release Conference call notes -- 2080826
Participants
Jones, Berkley, Riddle, Schultz, Ludaescher, Barseghian, Altintas, Brooks, Leinfelder, Crawl, Tao
Roadmap URL
Agenda and notes
1) Overview and discussion of roadmap
-- Chad provided overview of module reorganization
-- # reorganization-summary
-- core is very hard to refactor -- lots of dependencies
-- Chad to develop list of actors with proposed new module organization
-- could/should also have some higher-level suites for actors (e.g., essential, ptiny, domain-specific)
-- Build system-- David will fix remaining items in the 2.0 bug list
-- save kar and search will be worked by Aaron
-- documentation -- check the assigned bugs to find responsibilities
2) Overview and discussion of blocker and critical bugs
-- KAR versus module format
-- matt: they should be combined, serve the same purpose
-- aaron: KARs are mainly about metadata; no need to replicate the code every time; should be fast
-- same applies to jars
-- cxh: need to be able to bundle up what was used for an experiment: preservation and archival of workflows
-- chad: original use cases had 2 types of KARs: transport and archive
-- ilkay: also use KARs as transport for run results
-- ilkay: better to split the two systems, making it clear what each is used for
-- chad: a lot of work to merge the two -- definitely would push the release back a lot;
-- no need to merge now; can slowly merge the changes with new module releases
-- but still wants KAR files to be able to contain executable (jar/classes)
-- because we need a way for scientists to wrap a new actor and send it by just using Kepler
-- aaron: building new actors that contain source code means the scientist programmers should learn and use the build system or use an actor developer kit available from Kepler or as a standalone app
-- derik: using our svn system for all actors developed by scientists would be a problem
(sean: there is support for using other SVN repositories)
-- christopher: need to support users who aren't sw engineers who want to just apckage up simple actors
-- matt: need to document exactly what a module is and what a KAR is and how they relate
-- can a KAR go in a module, or vice versa?
-- what format should a run archive take?
-- what can/should go in each type of file?
-- what are scenarios of use (e.g., publication-ready archive)
-- how should projects that use Kepler make use of KARs and modules?
-- cxh: build system bug (ant taskjar)
-- cxh: test system needed for build system itself (will agree to punt)
-- matt: bigger problem is lack of tests for the system
-- derik: should develop manual list of tests to run through
-- aaron: might be better to have other devs test someone elses code
-- derik concurs -- get someone else to test their code, even if via the gui
3) Configuration system
-- matt: overviewed new ConfigurationManager
-- sean: should have one central place for all configuration
-- general agreement that a simple, central config manager would be helpful and not limiting to future extension architectures
-- ben: major issue is the 'cascade' of properties
-- aaron: modules can already do anything they want for configuration
-- a lot of properties in core shouldn't be settable by modules
-- really should be defining what extension points are
-- decision: pursue it, and chad will lead the architecture
-- will evaluate how much of existing config to refactor into new system; not all would need to be done for 2.0
4) Discussion of timing of release
-- A early fall release is the target
-- need to revisit as progress is made each week