Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.


  • Infrastructure
    • GIT!
      • George: really, just do it.
      • Darren: Stash integration with JIRA is nice.
        • but it doesn't bundle notifications
      • Will we do pull requests?
        • Only if it's the only mechanism.  Multiple mechanisms are hard.
      • Can we make it easier for contributors?
      • Do we do the Linus model for merging?
        • Paul: is it a developer's job to merge?
        • Sean: not a problem for merging between branches
        • Matt F: It isn't free.  There is work involved.
        • Downside of open stack approach: lots of moving parts
          • Leif: you can pick a subset of those parts
      • Action: finish evaluation

  • Testsuite
    • Dependencies are the hardest part.  Debian packagers:
      • attest doesn't work (but do we care?)
    • Paul: can we name the test suite?
    • George: trying to get all of the tests to run usually blows up.  The documentation is lacking.
    • Torey: versioning of the test suite?
      • Paul: I have a solution to that.  It's radical.  Remove the tests from the test suite (and put them somewhere else).
      • Version the 'stock' test suite libraries
      • Action: determine what it would take to do that as we move to git
      • Have test suite in docker.  Community: provider docker config


  • Simplifying sitting behind a proxy
    • Different identification schemes
    • Registration state dictates endpoint/device state
      • If I have a call from something, assume it exists
      • Have the presence of a call signal that an endpoint/device state exists.  Derive from the signalling.
      • Action: Daniel to provide an e-amil mail describing the problem
    • Need to know system load
      • NEED: CPU usage (Sean Bright and Malcolm: SNMP?)
      • Channel count
      • Memory usage
      • Torey: we send an INVITE once a minute and see what happens
      • Call quality statistics (secondary indicator)
      • Distribution of response times on requests/responses
      • Call quality (see above, again)
      • Darren: two scales - one for concurrent, one for calls per second
        • Caching mechanisms in Linux can be self-defeating.  Caching at that call volume can be rough
          • Look at IOStat
  • Simplifying routing/dialing of SIP URIs
  • Federation



GIT needs to happen. Need to improve our infrastructure and make it clearer how someone contributes to Asterisk.

Document features better, including scenarios. Make it easier to configure/deploy Asterisk.

Pushing Asterisk into acting more as a media applications server is very important.  It feeds into scalability, into making it do more than what it can do today, and affects more pointedly the features in ARI.