Implementing LOCAL_CREDS socket credentials

[Havoc Pennington]

Julio M. Merino Vidal wrote:
> 
> Prior to submitting this to the Bugzilla I would like to reach an
> agreement on the change because you probably want to avoid modifying
> the protocol at all costs...  At the moment I do not know how this
> could be done without the change, though... Fortunately dbus is still
> not 1.0 so it "can" be done (after all, the API was recently broken
> during an update, right? ;-)
> 

We could break it, there are ways to avoid it though I believe; one 
would be to use a different auth mechanism name like 
EXTERNAL_LOCAL_CREDS perhaps, though it would be a longer patch since 
it'd involve doing LOCAL_CREDS on a different codepath from the existing 
creds stuff.

The patch seems to write() in dbus_read_initialization_byte and read() 
in dbus_write_initialization_byte, that is somewhat confusing ;-)

Havoc