[systemd-devel] High level blurb -- pass two.

David Farning dfarning at gmail.com
Tue Mar 4 13:11:49 PST 2014


I wanted to take another pass at a brief intro blurb. It can take a
lot of passes to prune the vast amount of information in the wiki into
a few paragraphs that someone can quickly read.

As someone new to this space I keep coming back to two reoccurring themes:
1. Users expect systems to be dynamic. For example, as we go about our
lives networks come and go; a private wireless network at home, a
wireless VPN at work, various cellular data networks in between.

This requires an evolution of PID1. The model is not longer a static
init system which brings up the system to a predefined state. A better
model is that of a state machine. As resources or users come and go,
PID1 should bring up services necessary to bring the system into the
correct state.

2. The eternal monolithic vs micro kernel debate. The arguments for
and against systemd seem to be very similar to the debates for and
against. There are pros and cons of a monolithic design. There are
several theorists who posit that microkernels are better than
monolithic kernels. However, in practice it appears that in specif
situations a monolithic design seems to be easier to implement.

An advantage of this approach is that it splits the conversation into
two separate threads. The academic micro vs monolith question. The
practical, what advantages do the various PID1 implementations provide
question.

Please note, I am not trying to influence the design in any way, I
rather like it :) Instead, I am exploring how to communicate about the
project at a high level.


More information about the systemd-devel mailing list