[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Another system management tool to disappear.

On Wed, Sep 2, 2015 at 4:54 AM, T. J. Duchene <t.j.duchene@gmail.com> wrote:
> Sorry had an issue that caused a premature post before I could
> finish it.

Happens to all of us every now and then, I think.

> On Tue, 1 Sep 2015 23:11:51 +0900
> Joel Rees <joel.rees@gmail.com> wrote:
>> There is a difference between what I asked and what you're telling me.

And it remains so.

>> Simply tweaking and recompiling debian or redhat is not what I'm
>> asking about, although it can be tedious enough. Nor is building a
>> functioning gentoo really.

Please re-read that.

>> I'm asking if you have built an OS from scratch, including the
>> userland tools and apps, for a specific, non-trivial purpose.

And that.

> That depends.

Only if you can convince me to back off the bar you say I set too high.

I didn't set the bar. The people who say, "Compile it yourself if you
don't like it our way!" are they who set that bar. I'm not the one who
could have lowered that bar by putting systemd in a parallel

>  If you consider using LFS to be the only answer you will
> accept, then "No", since as I said, I have never used it.

Well, LFS is not the only option, but it's the easiest of the options
I see for "compiling it myself" relative to the current discussion.
Everything else requires more experience, skill, and time.

Steve mentioned (off-list, was it?) that an OS without systemd is
(relatively speaking) not that hard to switch a new init under, where
an OS with systemd is significantly more difficult. Perhaps thinking
about that should point you in the direction I'm thinking.

> If you
> consider that I have taken existing code, compiled, rearranged, or
> added to it to save time, then the answer might be "Yes."
> It really depends on if you accept that I have rebuilt most of Linux
> over multiple occasions, but never all at once.

As I say. this is not about your CV. It's about whether you understand
what "compile it yourself" means in this case.

> I suspect that anyone who has done so is a tiny minority on this list.

Not as few as you think, I'd say, but the relative minority numbers
are really rather to the point that "compile it yourself" is not an
argument in favor of systemd. Quite the opposite.

> I do not think that it is fair to judge what a person is saying based
> on that.

Once again, I am not talking about anyone's CV.


And so you repeat a bunch of that self-styled cabal's talking points.
I let you waste my time that direction last night. I've got to get
some sleep tonight.


     Build it yourself!

you say. Sure. Simple.

But cleaning out the bits and pieces of systemd api from various
packages takes time. Even if most of the packages have compile-time
switches, there is a lot of recompiling, and the current state of the
debian community does not allow for the community to do much testing
of the results. That means, really, that the DIYer has quite a bit of
regression tests to build, in addition to tweaking the build variables
and getting the compiles started on whatever passes as his or her
build farm.

The partial replacement for su that started this thread is a case in
point. Every package that uses it will be rather intimately tied to
systemd, and I, if I were to build a completely systemd-free debian,
would have to re-write parts of those packages, hunt for or write
replacements for some packages, and write new regression testing for
the results.

In addition to the tweaking, re-writing, collecting alternative tools,
etc., it would mean for me, several days, minimum, of down time during
the build process, since my build environment is my workstation, and
my CPUs are not fast enough to allow me to compile and do certain
parts of my regular work at the same time.

Working from Debian sources is going to take more regression testing
than going through an LFS recipe. If I go the LFS route, using
binaries from debian or fedora would be inviting coredumps. Using
source, less so, but then I end up re-compiling more non-OS apps.

Even working from Debian sources, stripping out systemd means subtly
altering the APIs and ABIs, inviting subtle bugs. Subtle bugs tend to
be the most dangerous.

Going one's own way on something like this essentially means
abandoning the community. And because of the amount of work required,
you end up needing to start your own community. So compiling it
yourself in this case requires more than technical skill and a little
extra time. It requires a charismatic leader to start a new community,

"Build it yourself." is only a whitewashed version of "Don't let the
door hit you on your way out."

Joel Rees

Be careful when you look at conspiracy.
Arm yourself with knowledge of yourself, as well:

Reply to: