RFC: D-Bus Overview Document

[Tom Cocagne]

> saying the protocol is "low-latency" (copied from the spec) isn't
> really accurate, what the spec should say is that it is designed to
> minimize round trips. The actual round trip _time_ isn't something
> dbus is optimized for.

I'd wondered how a central daemon with fairly expensive message
routing capabilities could claim low-latency. Certainly, in my own
implementation of a minimal DBus server I didn't even try ;-). The
minimized round-trips makes complete sense though. I'll update the
document accordingly.

>> The "hierarchy" implied the path structure is almost purely conventional
>
> it is used in libdbus to allow some objects to be "virtual" (you can
> handle a subtree of the path structure without actually creating an
> object instance for every path you want to support). it's also used in
> introspection iirc (you can introspect for a subtree). But you are
> right that it's up to the implementation whether to do anything
> meaningful with the path structure.

Yeah, this is something I found pretty confusing while developing
txdbus. It's mostly, but not entirely, conventional and, depending on
the implementation, the level of enforcement and support can vary
substantially. I tried shooting for a high-level, mostly-correct
description for the overview document but I'm not sure I hit my
target.

I like "virtual" objects idea and it's something I hadn't considered
for txdbus. I'm suffering pretty bad burnout on that project at the
moment but I'll keep it in mind for a potential future enhancement.

Thanks for the review Havoc.

Tom