• Cosmo automated client testing
• Cosmo 0.2.7 release
• Snarf
• Scooby release

Notes

Scooby is going well; some stuff taking longer than expected

• conversion of dates daunting; MDE can give a feel for problem & time scope
• QA: Selenium isn't going to work
  • JSUnit tests will work, Mikael is developing these for now and can check in
  • JSUnit tests will be checked into main Scooby trunk along with documentation
  • these tests will be transitioned from Mikael (QA) to Bobby and MDE (dev) after they're in place
  • Mikael will transition to higher level tests and investigate tools other than selenium

Cosmo: QA

• Litmus tests should be part of tinderbox -- with certain tests skipped or ignored!
  • But BCM says we're not ready -- not stable enough to run tests automatedly.
  • Goal to be ready for that in a week or so
• Heikki started throwing together "Silmut" using Python HTTPLib and ElementTree^? (will be part of Python 2.5)
  • uses Doctest so you don't even have to be so familiar with Python
  • the doctest spits out the results the first time you run a command so you know what to "expect"!
• Grant and Lisa were using Zanshin for the same kind of thing Heikki was using
• BCM suggests considering a Perl library-- but there are probably less people who can pick up perl than python
• Slide approach: data files for requests, expected responses
  • Lisa used something like this before, and even simple data files and basic test running engine can do 80% of the interesting cases
  • Grant investigated Slide's data-file-based engine
• Ideally this automated client testing could be run from JUnit or Maven if we expect developers to run it
  • we have more flexibility if we run it from Tinderbox, probably
• Heikki and Grant will send out examples of the doctest and Slide data-file approaches

Cosmo: 0.2.7

• Foxmarks/foxcloud found some critical data consistency issues around creating accounts simultaneously -- needed to put the home dir creation code into a "synchronize" block
• When the user hit "cancel" in foxcloud, dropping the cxn before uploading whole bookmarks file, Cosmo was storing the partial (corrupt) file. Solution was to require a Content-Length header, and Cosmo explicitly compares the value to the number of bytes read. This required changes in Jackrabbit too. Cosmo logs a 500 in this case.
  • When a proxy is in the road and it doesn't drop the connection, there's some way to know there's no data to read off the socket (???)
• 0.2.7 went to Foxcloud before OSAF QA ran acceptance
• OSAF QA will still run acceptance before we upgrade cosmo-demo
• Todd wrote scripts to use CMP and find and fix the broken accounts

Snarf

• Bundled distribution of Cosmo, Scooby, Tomcat, Jackrabbit
• Name picked just because BCM needed a name for that bundle -- a name for the download, for the SVN module, etc.
• Easier for developers to do a simple build process which doesn't require the server to be shutdown in order to iterate on the GUI
• Another option for users of Scooby is to download scooby only and deploy into their own app server
• Maybe it should be a sub-project of scooby?
• What about an overall server project, of which Scooby and Cosmo are both sub-projects? This might be less confusing than we had even before Snarf was created? Would it be good to combine the communities to a certain extent?
• We note that the developer expertise for Cosmo and that for Scooby is somewhat disjoint...
• BCM envisions adding java code to Snarf that's independent of Cosmo or Scooby -- e.g. code to manage/install Tomcat, or a unified admin console, or a JMX object schema for programmatic management of the entire server.
• What's the definition of "a project" at OSAF? At what point do you create a new project? How much overhead does a new project entail?
• BCM proposes re-unifying the server projects under one SVN, one mailing list, etc. He'll send this proposal to the list and ask for "+1s"

Misc

• Jared thinks that with top hardware FoxCloud^? can scale to 100,000 users with Cosmo. This is great for Cosmo at this point!
• Sheila wants to get some progress on Chandler synch performance in 0.7. BCM, Lisa, Grant, Morgen and Sheila need to get together in a room with some network traffic data to make what ought to be quick progress on this.