DBus over the network; clustered machines acting as one; service discovery

[Havoc Pennington]

Hi,

On Tue, Apr 8, 2008 at 10:04 PM, Shawn Rutledge
<shawn.t.rutledge at gmail.com> wrote:
>  So now I think after
>  having started on just one machine, eventually it will have to evolve
>  the other direction (going from one machine to clusters), and the end
>  result ought to be the same: it will do some or most of what Corba
>  did, and hopefully without the same cost.

FWIW, my hope is the answer is "no" - that dbus will not try to do
most or even a large part of what corba did, and that dbus will not
evolve to support clusters. CORBA was afaik created for the use case
where people might talk about "service oriented architecture" or web
services or something today. dbus is stupid for that.

As the dbus faq says, I simply don't believe that there is a one true
IPC system, that all IPC is the same, or that it's even _desirable_ to
use the same protocol in every case where a socket is involved. I
think trying to create the one true IPC system is nonsense. If you
have to make a system 10x more complex in order to support all
use-cases with one system, you are better off having multiple simple
systems tuned to individual use-cases. It's not like using sockets is
such rocket science that it must be carefully contained in only place.

Anyway. dbus is not the same thing as CORBA whatsoever, IMO. I think
CORBA absolutely did not provide the most important things that dbus
does provide: app lifecycle management/tracking, session scoping,
secure system bus allowing us to move between system and user session
security contexts, are some examples. The IPC part of dbus is almost
incidental.

Really, why would we try to make dbus good for clusters and
service-oriented architectures and whatever? There are already a ton
of things that do that, that people already use, that were designed
for that. And none of the people working on dbus even understand those
use-cases, afaik.

Do one thing well.

Havoc