On 1/7/2011 7:10 AM, Glen Gray wrote:
On 7 Jan 2011, at 14:46, Dave Neary wrote:
This raises then the obvious question of Why ?
What tangible benefits does moving to Upstart offer.
I can answer this...
Upstart supports old init style scripts, but that's not all it supports.
You can do things like have inter-service dependencies, so that if
service A needs service B to be available first, they you can tell it to
wait for service B in its upstart config file.
Upstart can watch config files and reload them, if it's told to, without
you having to do so explicitly (which is pretty cool).
Upstart supports on-demand starting of services (in a DBus like way).
systemd offers essentially the same benefits.
I understand the general benefits of Upstart. What I mean is what does it offer
MeeGo over SysVinit for the upcoming 1.2 release. Especially if the plan is
still to move to systemd for 1.3
(see Carsten's link to the Arch meeting). The MeeGo init script stack (at least for
x86) is based on the MoblinV2 stack. And the MoblinV2 stack was tailored to be fast
and use SysVInit. The premise for not using Upstart then was along the lines of
"Starting lots of bloated blocking code in parallel isn't as good as finding
out why things are slow and fixing them).
It seems like a futile effort to port to upstart for 1 release, ditch it and
move to systemd. The only potential benefit would be that the boot process
would be structured to depend on some of the parallelisms but that seems like
little incentive.
the only reason to do upstart now is the integration with the security
framework. If it wasn't a hard requirement for the security framework,
we'd not be doing this one-release weird choice.
_______________________________________________
MeeGo-dev mailing list
[email protected]
http://lists.meego.com/listinfo/meego-dev
