Proposal and RFC: DAL, the Desktop Abstraction Layer

[Havoc Pennington]

On Sat, 2005-01-15 at 14:57 -0800, Tim Rue wrote:
> Tried to catch up on this thread, but not sure I have.
> 
> But I want to say that dbus should be kept simple as reasonable possible that it
> might then become a standard and open/versatile IPC mechanisim/message system.

Don't worry about it. Virtually everything discussed in this thread has
already been addressed in dbus well before now.

> I do wonder if dbus might be heading towards a complexity beast itself, trying
> to handle more than it should.

No. As I've said a number of times, we aren't putting in whatever is at
issue here unless we get real-world experience to justify it.

I've been stripping out API recently in fact, punting more to the
bindings.

> Build the road, then figure out the traffic lights and traffic cops as needed in
>  use... for your system and don't impose on others on their system, for their
> traffic may be of a different nature.

I will give fair warning though that D-BUS has very specific design
goals; to be a session bus for the "mainstream" free desktop
architecture (GNOME/KDE-like architectures) and to be a systemwide bus
for free UNIX-like systems (primarily Linux, but if *BSD were interested
that too).

If people start asking to extend it in ways that are useful for web apps
or scientific clusters or other random IPC stuff, I'm just going to say
no to any extra complexity for those purposes.

Which is to say that D-BUS isn't supposed to be a "generic" IPC system,
it's supposed to do two specific things well. CORBA is there for those
who want an IPC system configurable to work in 1000 ways.

Havoc