dbus mini-summit at Desktop Summit 2011

[Colin Walters]

So Thiago, Will Thompson and I sat down for a few minutes at the
Desktop Summit here and talked about the current state of DBus as well
as the future.

I'll just paste my notes before my battery dies, we can elaborate this
on a wiki later:

* Land more of smcv's work
  - Release 1.6

* Thread Safety in libdbus
  - Probably permanently broken in libdbus

* Resource limits
  - Pretty much broken, but we need reliability

* Talks of DBUS 2.0
  - What is slow
    - Avoid the bus having to modify message headers as they go through
    - Avoid revalidation
    - Marshalling does revalidation and/or is unnecessarily slow
  - What do we need to break
    - floats
    - Possibly investigate unreliable signals
    - Wire format (talk to desrt)
    - Drop introspection in favor installing schemas
      - Or at least make it binary (GVariant)

* Signal connection
  - PropertyChanged signal not in Qt
    - Sender might be spammy
  - Know if other side is listening
  - Receiving a signal, knowing which sender sent it (which match rule matched)
    - Match rule returns a uint

* Object manager API
  - Implement in Qt?
  - Maybe need "I'm not listening anymore API"

* Maybe types
  - Feature negotiation
    - Toss SASL text format

* Investigate kernel multicast sockets

* Priority inheritance
  - High priority process like phone app doesn't actually work because
    bus daemon is lower
  - Thiago says that kernel futex has PI which