Issues while using DBUS over TCP

[Eric Jonas]

For starters, I owe you one (many?) beers. Thank you so much, I was just
sitting down to spend my entire day figuring this all out. 

> I was able to communicate with a DBUS session service running on my
> target platform using the modified dbus-daemon. I am sure some of my
> steps or explanations can be improved by someone with a better
> understanding of the DBUS architecture. I am open to any corrections
> or improvements and I hope this helps someone contemplating a similar
> use for DBUS.

I've confirmed that your simple modification to bus/bus.c works, and
lets me connect to the bus remotely with J5's excellent "d-feet" dbus
gui. I've not actually put anything on the bus, yet, but I wanted to get
my thanks out ASAP. 

Also, I'm of course very interested in the python bindings as well -- so
thank you for figuring that out too. 

One potentially vexing issue is that you don't appear to need to set
<auth>ANONYMOUS</auth> explicitly in the conf file for this to work. 

When you say : 
        
        The DBUS documentation on authentication (<auth>) is a little
        sketchy when it comes to ANONYMOUS authentication. Originally, I
        specified every form of authentication I would allow. What I
        found out is that the order in which these are listed is VERY
        important (and contrary to what the dbus-daemon documentation
        seems to indicate). This was apparent when using a Python client
        on the host computer. If I reversed the order in which the
        ANONYMOUS authentication is specified from first to last (e.g.
        <auth>ANONYMOUS</auth> comes first before the others) my Python
        client could not authenticate successfully. Finally, I removed
        (by commenting it out) all the <auth></auth> entries and my
        Python client can still authenticates anonymously. Obviously, I
        am missing some subtle point here that perhaps someone can point
        out.

I agree this might be an issue -- dbus devs, do you have any pointers on
how we might get around the "no <auth> entries ==> allows anonymous"
problem? It feels like users should have to really, really, really
specify that they want authentication disabled. 

Glenn, once again, great job and thanks!
			...Eric