Mailinglist Archive: opensuse (856 mails)

< Previous Next >
Re: [opensuse] Re: on something being monolithic and proprietary.
  • From: Todd Rme <toddrme2178@xxxxxxxxx>
  • Date: Sun, 23 Oct 2016 21:52:47 -0400
  • Message-id: <CADb7s=uJ8kJPHC1jcobo3bSjv2GSNZiFLjgtt5r6Ucv52Jn39A@mail.gmail.com>
On Sun, Oct 23, 2016 at 7:06 PM, Linda Walsh <suse@xxxxxxxxx> wrote:
ianseeks wrote:

On Friday, 7 October 2016 10:19:36 BST Philipp Thomas wrote:

* Larry Stotler (larrystotler@xxxxxxxxx) [20161002 22:50]:

systemd to me personally has always seemed to be a solution in search
of a problem.


Thats one of those statements i see where a lot negative and incorrect
ideas have run out, a bit like "its monolithic".

---
Caught this in looking for something.

You apparently didn't see where I explained why it was
monolithic, and I wanted to explain why it was -- as well as
using "proprietary" interfaces (ones owned by systemd vs.
non-proprietary interfaces that might be considered "open" or "public".

It is monolithic because its parts are not designed to be
"drop-in-replaceable" by any other non-sysD part.

Of course they are. It is just that no one has bothered to make such
replacements in most cases.

You can't replace any of the parts of systemd that have replaced
the earlier parts. Good example: syslog. Syslog was easily
interchangeable with ng-syslog and rsyslog. But none of those
logs are able to replace systemd's journal. You can add them on
as an afterthough -- but not as a drop-in replacement for journal.

There is an enormous difference between saying program A can't be
replaced with program B and saying program A can't be replaced at all.
You only provide examples of the first case, but then claim the
second. Yes, systemd's tools work in different ways than previous
tools. That is why people wrote new tools to begin with, to satisfy
needs and use-cases the older tools didn't fill. So it is no surprise
that older tools don't act as drop-in replacements. But there is
nothing stopping someone else from writing a tool that could act as a
replacement.

Say
I develop an init and want to use it in place of SysD -- but just
for capturing dead procs and such (creating a subscription mechanism
usable by other "parties", including SysD). Could I simply drop it
in and have it work?

Of course you could, as long as it provided the features needed by the
programs you want it to work with. systemd's components all use
standard linux tools to talk to each other and all talk using open,
documented interfaces. There is absolutely nothing whatsoever
preventing anyone from replacing any one of them with their own,
completely independent program. This hasn't happened much because
opponents of systemd aren't willing to put in the effort, there is
nothing whatsoever preventing it.

Of course not! That's why SysD is monolithic -- it can't be used
for its separate parts which are mostly indivisible as they use each other
in ways that are _proprietary_ to SysD (i.e. there
is no *open*, widely supported interface to interact with them.

The interfaces aren't "widely supported" solely because no one who
opposed systemd has made any effort to support them. There is nothing
stopping anyone else from supporting them other than the fact that
systemd opponents aren't willing to put in the effort.

--
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse+owner@xxxxxxxxxxxx

< Previous Next >
This Thread
Follow Ups