[jmspring]

I may be showing my gray ear hairs, but I remember the early days of Gentoo where ZOMG boot 3ms faster by optimizing! The init portion of systemd reminds me of those days.

Your mention of requirements for service execution is certainly an interesting avenue for me to think about.

Honestly for my day to day "spin up VM, do some experiments/coding, fine tune" it doesn't matter. For some other work I am doing dealing with longer term system maintenance/administration, I've been slow to adapt.

Thanks for the insights.

[the_why_of_y]

> The init portion of systemd reminds me of those days.

There is quite a big difference in that systemd was never designed to boot fast - it was designed first and foremost to boot correctly, and the much increased performance was mostly a happy accident of the design they came up with.

[falcolas]

This isn't backed up by Lennart's own blog post about the initial Systemd development:

> Unfortunately, the traditional SysV init system was not particularly fast.

> Another thing we can learn from the MacOS boot-up logic is that shell scripts are [...] slow in execution.

> it is also our plan to experiment with systemd not only for optimizing boot times

It seems like the real motivation was exactly reversed from your comment - speed first, and correctness as a happy accident.

He does put up a later post (biggest myths about systemd) some three years later which tries to change this impression, but that reads much more like an attempt to move the goalposts than an actual design goal.

http://0pointer.de/blog/projects/systemd.html

[the_why_of_y]

You have a point - clearly my comment over-stated it - let's say that both correctness and performance were design goals.

At the time Upstart was considered "the future" ... read the comment from Scott James Remnant on this post (unfortunately there doesn't seem to be a way to link to a G+ comment directly?):

https://plus.google.com/+KaySievers/posts/C3chC26khpq

The point is that Upstart's event model was found to be flawed, and the copyright assignment policy meant that fixing Upstart was undesirable.

This explains the filesystem problem (both relevant for reliability and performance) and claims that systemd solves it but Upstart cannot:

https://plus.google.com/+LennartPoetteringTheOneAndOnly/post...