On Wed, 14 Mar 2018 11:11:23 +0000
Arnt Gulbrandsen <arnt@???> wrote:

> Didier Kryn writes:
> >     Do you remember any of these comics where the driver of a 
> > car opens the motor to repair, throws away a bunch of parts, and 
> > then the engine starts again and the guy goes away with the car? 
> > Here we are with Linux. The BIG piece to remove was systemd, but 
> > there are quite a few others... follow my eyes.  

>
> Be careful that you don't end up one of those backseat drivers who
> explain afterwards that implementing this or that should've been
> simple because obviously blah blah. Those people are terribly
> annoying to developers who've spent weeks or months battling tricky
> issues, trying to make the code work in many cases, for many users.

As one such back seat driver, allow me to explain. When you've been
both programming and using for a long time, you get a feel for the many
ways something can be done, and when a developer who's spent weeks or
months battling the tricky issues presented by *a specific
implementation*, such as systemd, says it's hard, often you get the
feeling it's hard because they're barking up the wrong tree.

Example: The illustrious Lennart Poettering used autodetection of
plugged in USBs as a motive for socket-activation, the whooo-big
justification for systemd. Sounded bogus to me. Within 2 hours I had an
autodetector for usb thumb drives built, using inotifywait. Several
others, I think including Karl Hammar and Jude Nelson, and some others,
posted vastly improved versions on the DNG list.

Did we solve all the problems? Of course not. Did we devote the
programmerpower that was required to do systemd socket activation? Not
by 3 or 4 orders of magnitude.

SteveT