Re: Let's have a vote!

[Mart van de Wege <mvdwege@gmail.com>]

Steve Litt <slitt@troubleshooters.com> writes:

> On Thu, 18 Sep 2014 08:10:17 +0200
> Mart van de Wege <mvdwege@gmail.com> wrote:
>
>> Steve Litt <slitt@troubleshooters.com> writes:
>
>> > A) Twine and baling wire is better than monolithic entanglement.
>> 
>> Yeah, after this I'm really not going to take you seriously anymore.
>> >
>> > B) If you try Daemontools, you just might switch your view of which
>> > is twine and baling wire, 
>> 
>> Yeah, mixing sysvinit and daemontools, especially if you only do this
>> to try and duplicate systemd features, is twine and baling wire. You
>> might want to check your assumptions, you don't know what I use at
>> work. Here's a hint: I'm a sysadmin. 
>
> Oooooohhhh, a sysadmin, I'm impressed. You win the debate on that fact
> alone!
>
> And, because you're a sysadmin, you can be forgiven for not
> understanding the future consequences of software whose every component
> needs to know the business of every other component, as well as the
> components of all sorts of other programs that happen to be in its
> environment. 
>
Thank you for the gratuitous personal attack. Now you know why I don't
take you seriously at all. I only mentioned my job because apparently
*I* have to check my assumptions, but *you* are free to assume anything
about me at will, as you just did.

But for the peanut gallery: systemd needs to know the business of the
programs it starts, otherwise it can't do things like socket activation
and process monitoring. I am well aware of what it attempts to do, and I
agree with the systemd devs that it is the right idea, even if I have
minor reservations on the implementation.

> But just for fun, why don't you try Daemontools on a couple of daemons?

I run it in production. I'm well aware of its strengths, and its
weaknesses. It definitely solves the problem of process monitoring
better than bare SysV init; but that's a rather low bar.
>
> And as far as the twine and bailing wire aspect, Daemontools is so
> incredibly simple that you can document what you did with about 6 hours
> of writing, so anybody can follow in your footsteps. Really, you should
> try it.
>
You can't read. Look up there, I said it's twine and baling wire
*especially if you only do this to try and duplicate systemd features*.

Mart
-- 
"We will need a longer wall when the revolution comes."
    --- AJS, quoting an uncertain source.