Cosmo Engineering Meeting 25 Oct 2007

Agenda

• Cosmo 0.8
• Release Candidate churn
• Split experiment discussion (BCM)
• Branching 0.9

Notes

• Cosmo 0.8 discussion - we are set to release Cosmo 0.8. There is an interoperability issue between Cosmo 0.8 and iCal 3. The out of the box collection created by Cosmo includes spaces in the name, and therefore the CalDAV url. This causes iCal3 to fail. For now iCal3 users will need to create a new collection without spaces in the name. There was some disagreement over the best long term fix, so we'll be having a list thread about this issue.

• Release Candidate churn * a number of the issues in the Cosmo 0.8RC's were related to problems with the testing environments. Folks in QA are already at work for fixes for those problems. * there was some discussion as to whether we were spinning new RC's too quickly, without waiting for a full test run to complete - this looks like a definite improvement that we should make * Adam has a list of areas where he feels test coverage is weak. He will send that out and get feedback from developers on areas they feel is week. We will allocate some developer time (2wks, perhaps spread over 2 releases) to helping to broaden our test coverage * There is a large hole related to migration testing - plans for improving this will be included in Adam's message * We discussed testing features more thoroughly and having tests for features checked in along with those features. This sparked a tangent discussion about problems testing the new overlay functionality because key functions (create a collection, add an existing item to another collection) are not available from the FeedService^?, making it difficult to write javascript only tests for overlays.

• Split experiment discussion - Brian reported on the split experiment - this turned out to be way more complicated than anticipated due to Java packaging issues - lots of jars needing to go into the global Tomcat library/class path, need to manage Hibernate resources like caches between modules, and so forth. Brian proposed backing off the aggresive split work that had been done and just working to break the UI out separately. The major barrier to this is the account brower's reliance on the Sevice APIs. He proposed two alternatives: * AJAXify the account browser (est 3-4wks work) * Leave the account browser UI as it is, and instead replace the Spring level controller for the account browser with one which speaks to the rest of the server using Atom instead of the service API Jared (and Ted) was concerned about the impact on performance - this ought to be localized to account browser, but side-by-side benchmarking ought to tell us more. Jared also wanted some more time to noodle on the impact to deployment scenarios, he promised to post concerns within 2 days, otherwise we will probably proceed.

• Branching 0.9 - we will branch for 0.9

Status

Ted

• Progress
  • Evangelism project
  • Website metrics
  • Cosmo 0.8 rc spins
• Plan
  • Cosmo 0.8
  • Cosmo 0.9
  • Cosmo 0.10 plan
  • More evangelism project

Brian

• Progress
• Plan

Randy

• Progress
• Plan

Bobby

• Progress
• Plan

Matthew

• Progress
• Plan

Bear

• Progress
• Plan

Jared

• Progress
• Plans

Travis

• Progress
• Plans

-- TedLeung - 24 Oct 2007