Platform Team Meeting 16 May 2006

Agenda

• Question of the week: How did sprint week go? What should we do more of/less of/same as next time?
• Alpha3 -- next steps
  • background sync
  • performance
  • stamping/domain model
  • sharing format
  • dump and reload

Notes

• Question of the week: How did sprint week go? More of/less of/same as?
  • same as: 2 days + list of tasks on whiteboard worked well
  • same as: in formal conversations, debug things with other people, have some unscheduled time
  • everyone in a room for a few days every so often a good thing
  • would be better if not at the same time as a release
  • liked the layout of the room
  • morgen wishes he could be there more than one day
  • less of: baseball game (not the greatest use of time, took up big chunk of the day)
  • more of: spread out social stuff at night
  • more of: timing apart from the release
  • more of: schedule so that phillip and morgen being there
• Alpha3 topics
  • background sync
    • bugs that are not reproducable, need help testing and come up with reproducable cases, no way to debug it
    • merge branch with trunk as soon as alpha2 is out the door? fix two stacktraces and then merge
    • andi will check in new merging as soon as the tree is open
    • time branch landing with epydoc
  • performance
    • write tests and set targets
    • mikeal is working on a data set
    • mikeal has a test framework branch, test the changes they are making to the unit test framework
    • when they merge, will impact functional tests
    • the two queries: bequeath to feature -- bugzilla task
    • sidebar collection mess -- schedule a review: talk to philippe -- jeffrey/alec, brainstorm some options
  • stamping/domain model
    • grant is the driver
    • ted to hand off issues to grant
  • sharing format
    • make the proposal more concrete: offers some opportunities for reuse, two frameworks have a lot in common
    • purse having a combined format -- combined information model
    • might be different on the wire format (different goals) -- understanding a common information model
    • code for upgrading and transforming in and out of flattened aspect -- best payoff (xml or pickles)
    • partition data: collections/parcels/repository?
      • set of named tables -- map back into data (how that happens determined by parcels)
      • table: collection of tuples of flat values
      • someone has to read data that there is no longer a corresponding class or kind
      • parcel has to be responsible for how name maps back into something
      • defaults will work reasonably, should all just work
      • parcels can define handlers, this table process this chunk of code
      • old parcels? (yes those as well)
      • many to many relationships
    • morgen to go back and think about requirements
  • "dump and reload" (application level)
    • how does this connect to install and the ui? command line tool? startup dialog? (yes)
    • startup dialog runs with no open repository -- standalone tool, must be headless/block free
    • separate executable?
    • if during install process -- people run the installer again, flagged as fix tool, some other tool, you may not have everything that you expect in the chandler/python environment, don't make the tool (startup options dialog) so that they depend upon chandler resources, you may not have the ability to load a chandler python -- berkeley db repair tools are not based on python
  • combined information model
    • elementary data types
    • how to structure the names for information sets
    • requirements for sharing format: read by cosmo
    • pje to get clear on what the other requirements entail
    • anything going on there that is helpful for the dump and reload
    • chandler fetch a resource from cosmo, file will have records that span the tables, each record processed by different part of the code, relate to attributes of different things
    • resource represents one thing? (resource is an item cloud -- may be several items)
    • support pre or post processing -- changes that span complex types (contact, contact name)
    • small subset of repository, querying operations, issue doesn't apply -- reload of entire repository and completely change which kinds do what. Doing that in sharing is much simpler -- modestly sized shares, do it in memory or don't store. Can we use same machinery for both situations when there is a schema change involved.

Status

Katie

• Progress
  • sprint week
  • dev list summary
  • sharing format writeup and meeting
  • stamping proposal and adhoc meeting (no notes)
  • wrote a few tests at the sprint
  • collection code review
  • cpia meeting & decisions
  • email discussion
  • alpha3 plans overview with Mitch
• Plan
  • wrap up vision elements prioritization, get target user info out there
  • next steps for alpha3 projects
    • performance -- assign tasks for new perf tests
  • schedule collections meeting with alec & cosmo/scooby folks before alec leaves
  • start dogfooding alpha2

Andi

• http://kumu.osafoundation.org/~vajda/logs.cgi

PJE

• Progress
  • Setuptools 0.6b1 released
  • Dump/reload discussions w/Andi
  • Posted pre-proposal info for dump/reload
  • Misc. developer assistance in IRC and email
• Plan
  • Final push for Python 2.5 projects
  • More detailed dump/reload work

Grant

• Progress
  • Attended a bunch of sprint week meetings (and a baseball game!).
  • Checked in fix for: 5769 Fiddling with recurring events can make them not show up
  • Wrote a couple of functional tests during the testing sprints: test of “Visible Hours” in the calendar, and tests that “New Item” respects the selection in the app bar.
  • Discussed with HeikkiToivonen and RobinDunn, and fixed: 5876 Can't emulate menu keyboard accelerators
  • Spent some time with RobinDunn figuring out what was going on in 5219 (Paste problem in Cocoa dialogs in 0.7α2). Worked with JohnAnderson to review and test the fix.
  • Gave a (probably not so comprehensible) lightning talk on writing setuptools-based Chandler plugin parcels.
  • Fixed some issues with ChandlerSyncServices (mainly resulting from changes in the Collection API).
  • Took a first cut at adding CalDAV free-busy-query REPORT to zanshin. Also, changed zanshin to use setuptools.
  • Figured out fixes (post-0.7α2) for a couple of other Mac menu bugs (5247, 5400).
• Plan
  • 0.7α3 recurrence work.
  • Catch up on chandler-dev@ email.

Morgen

• Progress
  • Background sync code clean up, including removing an unrequired repository view
  • Re-enable the username/password dialog for subscribing to shares with no matching account
  • Background sync testing
  • Fixed a couple trunk bugs
• Plan
  • Give the user the ability to change the sync interval (or go 'manual only')
  • Clean-up new sharing code

Ted

• Progress
  • Sprint week
    • Wrote a unit test for Scooby's RPC service, patch attached to [Bug:5911]
    • Collections, CPIA, Contacts, Portal, Sharing Format meetings
    • DCamp - brief summary at http://www.sauria.com/blog/2006/05/15#1513
• Plan
  • Assign Google Summer of Code mentors
  • Committer Policy changes
  • Get ready to have discussion on portal project