Re: how to remove libsystemd0 from a live-running debian desktop system
On Tue, Feb 17, 2015 at 11:52:21PM +0000, Luke Kenneth Casson Leighton wrote:
> On Tue, Feb 17, 2015 at 10:52 PM, Josh Triplett <josh@joshtriplett.org> wrote:
> > So, please go educate yourself on what libsystemd0 actually does,
>
> i know what it does, and what it does - technically - is *not* the
> issue that i am concerned about.
Then you've given no reason whatsoever for anyone to remove it. Because
without that, all the same arguments apply to how hard it would be to
remove PAM, glibc, glib, or many other libraries, most of which would be
even harder to remove than libsystemd0.
>From current unstable:
~$ aptitude search '?depends(libsystemd0)' -F '%p' | wc -l
70
If you're going to artificially inflate that number by talking recursive
reverse-dependencies, then I would suggest that you look at the
number of packages removed by attempting to remove libpam0g, or
libglib2.0-0, or libc6. By any reasonable definition, those are not
"optional" either.
> > and if
> > for some reason you still consider it a problem after doing so, you'll
> > need to explain why,
>
> i have done so, a number of times. take away the name of the
> library. take away what it does. take away how it does it, because
> none of those things are relevant.
Apparently they are, to you, or you'd be making the same rant about
something that didn't contain the substring "systemd".
> what *does* concern me is that it takes such incredible (and amazing)
> efforts by people like adam for the average end-user or sysadmin to
> contemplate replacing {insert nameless package}.
And you've now completely ignored my question. If you want people to
*care* that it's hard to remove libsystemd0, you'll have to explain why
anyone would *want* to remove libsystemd0. There exist a pile of other
libraries in Debian that have far more reverse dependencies and would
take much more work to remove.
> that *is* the problem. i'm aware that there are many people in key
> positions in debian who do not see this lack of choice as being the
> problem, but i can assure you that it is.
Your assertions have little to no credibility; even less every time you
write a new mail in this thread. I've already seen evidence (in the
form of mails in this very thread) demonstrating that people who have
said in the past they care about preserving non-systemd init systems
(e.g. Russ) do *not* see the point of avoiding dependencies on
libsystemd0. So you're currently arguing "on behalf" of a niche of a
niche, and you're doing absolutely nothing to explain why the subset of
users who don't want libsystemd0 installed (which thus far appears to
contain 2-3 users) is worth a *disproportionate* amount of development
time.
> > because as demonstrated in this thread, even those
> > developers in Debian who still do care about non-systemd systems do not
> > agree with you that it's a problem. See, for instance, Russ's response,
> > which you lauded while failing to actually comprehend, since you seem to
> > believe that his response described something that needed changing
> > rather than describing the current state.
>
> i believe tiredness may be affecting my ability to understand the
> point you're trying to make, here. i'm genuinely pleased that russ
> (and adam) came up with the same possible solution (dynamic library
> loading) that, if deployed, would end this entire issue because it
> would allow people to make a choice.
If you're (intentionally or unintentionally) misinterpreting Russ's mail
as suggesting that dlopen would be a good idea, let's stop talking past
each other: no, that'd be a bad idea, since the amount of additional
code needed to do so would make the use of libsystemd0 more painful,
when libsystemd0 is not a sufficiently large or invasive library to
warrant that. libsystemd0 exists to avoid having to write duplicate
code in numerous programs; adding more such duplicate code to use dlopen
would defeat the purpose. We use dlopen for libraries that are either
1) absurdly huge, 2) have unwanted dependencies, or 3) have issues that
prevent depending on them (such as libdvdcss).
libsystemd0 itself is specifically designed to not depend on any other
component of systemd; libsystemd0 *is* the thing you link to to
*optionally* use systemd functionality. And you still have yet to
explain why anyone would need to remove it. (Arguments against systemd,
credible or otherwise, do not automatically apply to libsystemd0.)
I'm all out of patience now, and I no longer have any hope that you
actually care about being taken seriously. I have no plans to respond
to any future mails from you.
If you *do* end up caring about being taken seriously, I would suggest
that you re-read my previous response, and take some action in response
to it *other* than replying, ideally involving 1) some research and 2)
some understanding of basic mailing list etiquette. Consider in
particular the mental models that other people might use to classify the
usefulness and credibility of your mail; if there were a
spamassassin-style mail classifier for where a mail falls on the scale
from "valuable contribution" to "crank", the useful value of a mail,
you'd be tripping most of the tests towards crankitude (several of which
I enumerated in my previous mail).
(https://en.wikipedia.org/wiki/Crank_%28person%29 is not intended as a
checklist.) One of the best pieces of advice I've ever seen regarding
mailing lists (and one I still endeavor to follow today): one writer,
thousands of readers, optimize accordingly.
- Josh Triplett
Reply to: