Starting the kdbus discussions

[Glenn Schmottlach]

>
> I don't see it that way.  I see this as a question of whether the DBus
> community wants DBus to become a part of systemd (for all practical purposes
> it would be).  While that has vendor implications as different vendors ship
> different parts of systemd, I don't think it's fundamentally a vendor issue.
>
> If I'm the only one that is interested in DBus being independent, I'm happy
> to shut up.  What I see currently is no one discussing it (though it is a
> typical time that many people have holidays, so they may be not checking
> e-mail, which is a perfectly valid thing to do).
>

I tend to agree with Ted. I primary work on embedded Linux platforms
(automotive IVI specifically) and regardless of GENIVI's perspective
on "systemd" support,  in this space having systemd available is *not*
a given. In fact, a custom init.c/systemd/upstart etc... tailored for
a specific domain will usually perform better because the target
audience is smaller and the use-cases are more constrained than a
general desktop environment.

With that said, I am certainly interested in kdbus - especially if it
offers performance improvements over vanilla D-Bus. So for my use-case
(automotive) I'm particularly sensitive to another dependency
(systemd) where one didn't exist before. In fact, due to the minimal
dependencies of D-Bus, I've been able to use it on Linux and QNX
platforms as an application bus for automotive use-cases where Linux
is not the most popular OS. I realize that nothing comes for free and
kdbus likely provides additional features/performance that may not be
easily portable to other platforms or OSes (given the obvious kernel
dependency). Certainly, if it is indeed true that kdbus is dependent
on systemd then my interest is diminished. Of course, I have zero
influence in any major distro like Redhat or Ubuntu and it's unlikely
my opinion amounts to much. It seems, however, that if systemd is
indeed a requirement to use this new D-Bus variant/replacement then
certainly assuming it will become the defacto replacement for D-Bus
(as it exists today) may rub people the wrong way. But hey, it's open
source, right? Do what you gotta do I guess but don't be surprised if
there is some resistance.

With that said, please confirm my understanding of kdbus:

1) Requires systemd integration in your distro?

2) Will require language bindings to libdbus to change . . . or is
this the shim the guys at Samsung are working on?

I ask about #2 because I have a Lua binding to D-Bus that uses the 'C'
reference library as it's glue to D-Bus. I was hoping the language
binding would *not* have to dramatically change to accommodate kdbus.

Thanks. . .