[arch-general] When will Arch switch to Upstart

Yaro Kasear yaro at marupa.net
Wed Jan 19 16:33:45 EST 2011

On Wednesday, January 19, 2011 02:25:28 pm Isaac Dupree wrote:
> On 01/19/11 14:03, Yaro Kasear wrote:
> > And comments about Ubuntu and their competence are entirely relevant to
> > this discussion, as Upstart is entirely their creation. Would you rather
> > I talk about people who had nothing to do with its code? The Ubuntu devs
> > are behind Upstart, they're not that great at what they do when it comes
> > to the actual system side of Ubuntu. Therefore why should we consider
> > Upstart an improvement.
> Argument ad hominem.

This was not argument ad hominem. I didn't call you names or insult you. Learn 
what it means before using that phrase. 

> We can be precise; it's more obviously rude that
> way.  Scott James Remnant wrote Upstart.

For Ubuntu. As an Ubuntu developer. Why do you think upstart's main page is on 
the Ubuntu web site, and not just Launchpad?

> I can't speak for Ubuntu, but
> I've seen Remnant presenting and he seemed quite competent.  Software is
> hard; Upstart was the first attempt at changing 'init' in decades, so
> there was little experiential knowledge of Linux 'init' development when
> it started in 2006. 

And that leads to my next question: What makes Upstart necessary?

Nothing. Boot speed is a trivial reason to overhaul the way UNIX/Linux boots, 
especially for an init system that is needlessly complex and far less 
accessible than what we already had. If we switch to upstart, simply beign 
able to edit a file like /etc/rc.conf will be gone, and setting up daemons 
"The Arch Way" would become unnecessarily difficult.

The fact that init hasn't been changed in "decades" like you claim is because 
nobody worth their UNIX development skill felt init actually NEEDED a change. 
It still doesn't. Just because you CAN change it doesn't mean you NEED to 
change it. SysV Init is fine for Arch, maybe systemd might be a pleasant 
change, but it's not a necessary change either.

> In fact, in the process of writing Upstart, Remnant
> and his co-workers made Ubuntu boot faster largely by working with Xorg
> and Linux kernel developers.

Boot times are an incredibly trivial reason to upheave the actual system 
process of a system like Arch. Rarely does someone actually need to get to 
their desktop in such a hurry, and it's far less important than the system 
being stable or fast during actual runtime. Upstart moves things in the 
opposite direction of that.

> There are now upstream changes due to the
> risk Ubuntu took with Upstart.  *Arch* therefore now boots faster
> because of Remnant.

First I ever heard of Upstart being important enough for something MUCH bigger 
and MUCH more important changing how they work JUST for the sake of something 
MUCH smaller and MUCH less important to work correctly. I'm going to have to 
see official upstream patches or it didn't happen.

> He's a pretty smart guy who knows what he's doing
> even if some of us disagree with what he's doing; I was at his
> presentation "How We Made Ubuntu Boot Faster"
> http://events.linuxfoundation.org/linuxcon2010/remnant

Again, booting faster is a luxury. People talk about it as if its very 
important, but it's trivial compared to a system running and running well, two 
things Upstart's known for not promoting.

> That is equally no reason to switch to Upstart.

There is NO reason to switch to Upstart at all, except that the OP seems to 
think that just because Ubuntu and Fedora did it, we should too, and plenty of 
reasons NOT to switch to Upstart.

> We can be grateful to
> Remnant and choose the best (technically & socially) solution *for Arch*
> *in 2011*.  Of course he's enthusiastic about Upstart but I'm sure he
> wouldn't mind.

Again, Remnant made WHAT contributions to Arch to date? You plan on pulling 
him from his place as an Ubuntu dev where mediocre code is praised to make 
Arch run worse just for the sake of a couple seconds less boot time? Here's an 
idea, how about we just rewrute the initscripts to be more efficient. That'd 
be much less of a trouble.

> (I don't pretend to know which solution this is, though
> it sounds like Arch's current init system, or systemd, are likely to be
> default in the next year or two.)

Sounds like you already have done this. Blindly singing the praises of an 
Ubuntu developer just because he is a good public speaker and that he "seems 
competent." Have you ever even tried configuring upstart manually? It fights 
you. Every step of the way. IT takes away that much flexibility from the 
system just for the sake of taking off a few seconds of boot time.

I have no opinion on systemd as an init system, as I have no experience with 
it. But I have used upstart enough to know that it'll take more than its 
developer acting like he's a good programmer for me to take it seriously.

> After writing the above, I checked my assumptions and Google found me
> Remnant's entirely reasonable blog post about systemd.
> http://netsplit.com/2010/04/30/on-systemd/

I'm talking about Upstart, not systemd. I'm not against systemd in any way, 
though I will have to ask what good reason we actually have for changing how 
Arch boots beyond "fast boots." I see none.

I am still unconvinced that Upstart is anything but a stupid idea for Arch, 
even if we're definitely not using it by default.
> -Isaac

More information about the arch-general mailing list