On 09/03/2019 09:28 AM, Bruce Hill via lfs-dev wrote:
On Tue, Sep 03, 2019 at 12:39:18PM +0800, Kevin Buckley via lfs-dev wrote:

Actually, as one of the people who strived to get away from the
default package ordering (more or less alphabetcial once a core
set of packages had been installed) when the SysV and SysD
books started to diverge, yes I am aware of that.
My original comment was mainly about LFS chapter 5, where
mandatory building order changed...
Which, IMHO, should never occur as package layering is stable and known.
(Here, LFS, by changing/finding new order, is responding to package
changing their dependencies, not a critic about LFS team)

This seemed an open opportunity for one who has worked on distributions using
"SysV-style init" (from UNIX System V), BSD-style, OpenRC, and now systemd, to
post a couple of historical links [1,2]. Not just used, but tested, built, and
maintained packages for 2 distros. When systemd was first announced, the BDFL
taught me to hate it. When Linus went off on Lennart and Kay on LKML, I hoped
systemd would die as did HAL. The reason it's caused so much strife has less
to do with the design of systemd, I think. Rather, the fact that many people
have used nothing other than SysV init (which, apart from the inittab
handling, is just awful) and been contemplating their navels.
;)
interesting comment, I strongly (as many) object on systemd
design itself, but lets say I am "contemplating my navel", please
help me to have a better systemd understanding.

As sysadmin:
- systemd is adding a layer of complexity, such, in cause of
  emergency/trouble, tools given by systemd doesn't allow to have
  contextual view of the problem. (A critical server is not rebooting,
  second counts!, why it not rebooting??!!, you need to resolve
  issue ASAP).
- Systemd (by design) doesn't allow me to proceed in CLI mode,
  working, starting process in a new/debug order, trying to
  have a clear view of what is working, Not speaking of
  systemd logs (a recurring critic about systemd logs).

Systemd paradox, systemd is a sys-admin tools, not designed
by a sys-admin.

As designer,
- The systemd original goal was to speed up booting process
  (a none issue in server context), to achieve this, project
  end up to annex many many system functions
  (udev, time, dhcp, etc...) to have them embedded within
   systemd...Now, systemd is taking a lot of
  responsibilities/decisions, being de facto a kernel over
  a kernel  (bad design?).
- Last time I checked systemd, main program was over 6000 lines (in C)
  without even one comments, hardly open-sources or documented....
  (as I am saying to younger colleagues, do not "piss" code,
   write something, yourself will be, still, able to easily
   understand in 6 month from now...).


About sysvinit:
I find it extremely simple, if you find it "awfull" it is,
may be, you tried to use it in a way it is not designed to
be used... (layering position and design issue)

:)
An the winner is:
RedHat, being RedHat, I would push all my weight and
power to have systemd widely accepted.
- Easier to sell support.
- Better to implement and support Linux desktop/tablette ...
  (but, AFIK and amazingly enough, the Redhat real market
   is the server one...)


So please, LFS team, keep sysV-style book,  as a teaching tool,
it allow a better transparency about booting component relationship.


I'll take OpenRC over SysV-style init any day of the week, though. SysV is a
nightmare. But now that my mind is open to systemd, it has been surprising to
find that everything in systemd is properly documented.

Kudos to Version 9.0, and your diligent work on this release.

[1] http://0pointer.de/blog/projects/systemd.html
[2]  http://0pointer.de/blog/projects/systemd-for-admins-1.html

Bruce



--
seen "Linux from scratch" and looking for ISO files
www.osukiss.org

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to