Discussion Topics

• Logging

  • Options

    • Use posix logger

    • Use message bus

    • Write to well known path

    • stdout / stderr

  • Issues

    • Accessing logs for debugging

    • Creating log entries during start-up

  • General consensus is to use posix logging (e.g. writing to /dev/log, using the standard system calls) (to be clear we are not specifying which system logging daemon a platform chooses to use)

  • Leaves open how do logs get exposed to the management system.

  • Do we need to define anything with respect to logging rotation or a dedicated log partition or TLS for log transport? These are recommended, but up to the implementer.

  • Side note: Should we consider a profile model for supporting different classes of devices

  • What do we want to do w/r/t facilities and labeling things based on application ID? [This is one advantage of using a message bus. Eloi suggests doing an experiment on what is possible to see what kind of filtering makes sense. Eloi is willing to help. Michael will work on a formal proposal. ]

  • Need to define the application ID. Norm notes that reverse DNS could be used similar to what has been done in the Java space (e.g. org.geisa . . . ) This may be challenging on some networks. If we use this, we may want a compact form. The advantage is that it avoids the need for a central authority. An option would be for GEISA to maintain a list, though this is centralized. Consider the IANA model. Norm to look at how the mobile world approaches this and come back with some thoughts considering the network limitations.

• Shared Responsibility Model