Re: in-kernel messaging (was Re: brasero requires gvfs)
On Thu, 4 Sep 2014 12:06:29 -0300
Henrique de Moraes Holschuh <hmh@debian.org> wrote:
> https://lwn.net/Articles/405346/
> https://lwn.net/Articles/484203/
> https://lwn.net/Articles/580194/
> https://lwn.net/Articles/537017/
> https://lwn.net/Articles/551969/
Thanks for these very interesting links.
> Be sure to read the comments as well.
Not time now, but I bookmarked them for further readings.
> Note that this won't cover the design of DBUS itself. But keep in
> mind that some of the modern uses of DBUS are very unlikely to be
> compatible with its original design goals.
One of these raised a question: how to make dbus passing large
amounts of data now?
From what I know, this is a very wrong question, you don't ask
an everyday car to win the 24H du Mans…
Not to mention a different moto in avionic than in Linux,
Linux says: don't touch what's working well, avionics says:
never ever touch what's working well (there are of course,
exceptions (think 1st launch of Ariane-V), but these are
exceptions that confirm the rule).
> Which doesn't mean its original design was stellar. DBUS does stink
> of NIH syndrome to high heaven, regardless of whether it deserves it
> or not. Still, it is necessary to look at what were the in-use
> alternatives at the time: I have this hunch that people flocked to
> DBUS because they were running away from the likes of CORBA.
CORBA serves a different goal, where speed wasn't involve at all
from its genesis; it was also only a design, which lead to some
odd proprietary adds as the standard did not evolve fast enough.
And to come back on "large data transmission", may be this could
be discussed between main IT actors (SW and HW) and chipsets (CPU?)
designers, as volumes are very high (just a thought).
Reply to: