[kwlug-disc] What is all this about systemd?

[Ralph Janke]

On 2014-07-19 20:44, unsolicited wrote:
> On 14-07-18 09:46 PM, Ralph Janke wrote:
> .
> .
> .
>> I am not convinced a monolithic kernel is the best solution
> 
> This is why I was quite fascinated when RIM chose QNX. I still await
> to see how that all plays out. True multitasking with only the kernel
> pieces necessary for any given install.
> 
> 'course the IBM / Apple enterprise announcement gives one pause on
> that. If Apple 'just works', (PDA) end users only need simple things
> to work sufficiently well, and IBM delivers what end to end enterprise
> is willing to pay for - sufficiently good may well again beat out
> technically superior.
> 
> 't will be interesting to watch. But I expect IBM/Apple will win, for
> the same reasons Ubuntu has.
> 

It depends what is meant by "true multitasking". At one point or another 
there always
may be resources that need co-operation or scheduling. However, the 
multi-tasking/
multi-threating problem is IMHO mostly independent from the issue how 
monolithic
a kernel needs to be. Monolithic has for me more to do with the "Single 
Responsibility
Principle" and general decoupling. I think there are less arguments 
these principles
have tremendous advantages in regards to the development process and the 
incrementing
of technical debt, however about the efficiency during run-time are more 
arguments.

'Just works' multitasking in general usually means highly independent 
processes/threats.
As soon as dependencies exist, it does not just work, but must either be 
specially
developed for it, or needs analytics that can make proper decisions. 
Regardless, most
of the time there is still resource blocking (mot likely memory access 
or i/o).

> 
> Re: Desktops
> 
> ...
> 
> The world is a world of users who's focus is not on the nifty things
> one can do with a computer, but with doing what they are required to
> do with as little mindshare as possible, so they can apply greater
> focus and time on whatever scratches their itch. Which isn't a nifty
> desktop widget or PDA app.
> 

I believe the idea of a lot of the FLOS people and "merging markets" 
proponents
like Canonical is that a full desktop is not necessary for the 
functionality
you describe here. Hence the desktop is stripped down to very 
rudimentary tasks.
Especially on mobile devices with small screens, this may be an 
interesting approach.

However, I have always argued that this is not my use case. My time is 
spent mostly on
development and not "playing and being entertained" by my desktop. I see 
the business
case, that's why I have full understanding why Canonical, Google & 
others have decided
to move in such a direction. And if they believe a lot of the 
unnecessary functionality
should be in the kernel, there needs to be a practical solution for a 
fork with
hopefully minimised additional maintenance costs.

-- 
txwikinger

Long live free/libre software