[mdlug] Systemd Should Be A Fork -- Why Is It Not?

[Adam Tauno Williams]

On Tue, 2015-09-01 at 15:24 -0400, A. Zimmer wrote:
> On Tue, 01 Sep 2015 13:13:45 -0400
> Adam Tauno Williams <awilliam at whitemice.org> wrote:
> 
> > 
> > There will be forks if there is someone with sufficient skill who
> > actually does *not* want things like working printer discovery.
> The multi-user, networked scenario does not apply to everyone.
> In particular, the single desktop/workstation operator who only
> has a connection to the Internet through an ISP would be severely
> burdened by such a scenario.

"burdened", how so?  By a sleeping process that never awakes.

And I was speaking of a physically connected USB printer.  Anyway, the
unnetworked device - even if only to a local network - is a very rare
thing.   With modern connective technologies be they 'network' or local
[USB, FW, Bluetooth, zigby, CAN, etc...] you need an event driven
scheme.  All devices, down to your phone, are complex and dynamic.

You build reliable and scalable [up or down] systems by messaging and
dependency injection.  This is always, eventually, true.

> Fortunately, up to now, Linux has been nearly infinitely configurable.

Yes, it has.  And the need to always configure it - which is kind of the
point.  Better is a system that configures itself for common needs.

> I don't need discovery.

Yes, you do.

>   I already know what is on my system and
> I can configure accordingly using the traditional static devices.

You can configure it to statically communicate with USB devices?  I
doubt your modern system is doing that.  UDEV is kicking in and helping
you out, dynamically.

> I write my own very simple init scripts and boot directly into a
> bash shell.  There is no need for security or hardening because I am
> the only operator and I have no open ports. 

Your understanding of security is flawed; controlling egress is as
important as controlling ingress [and you do have ingress as you open
socket connections and retrieve data[.