:: Re: [DNG] /usr to merge or not to m…
Pàgina inicial
Delete this message
Reply to this message
Autor: Alessandro Selli
Data:  
A: dng
Assumpte: Re: [DNG] /usr to merge or not to merge... that is the question
On 23/11/18 at 14:32, Stephan Seitz wrote:
> On Fr, Nov 23, 2018 at 01:09:17 +0100, Alessandro Selli wrote:
>>   I do get the reasons the merge proponents prefer this filesystem
>> layout.  What I rant against is their choices being imposed on me.
>
> They are not imposed on you. No one is going to your desktop/server
> and is changing your disc layout.



  Debian is (going to).  In order to avoid this imposition in Devuan too
the install procedure needs to be modified to provide with a choice and
a new package is beeing introduced.  Debian is not going to do this,
because to them the split /usr is:


1. silly;
2. old-time;
3. useless;
4. hard to maintain though it's been done since the early 70s;
5. not datacenter and cluster friendly;
6. esoteric.


>>   While I did not state this claim, I thought desktop use was one of the
>> targets Devuan would try to accomodate.  If this is not the case then I
>> think I'd better go search some desktop-friendly distribution.
>
> Sorry, you’re funny. A desktop-friendly distribution is certainly not
> a distribution which will ask you about your partition layout and
> filesystem choices because the majority of people who want to use a
> desktop-friendly distribution is not interested in these details.



  Speak for yourself.  "Desktop install" does not equate to "directed to
the uncultivated masses", even if this was the most common case.  Lots
of long-time Unix and GNU/linux sysadmins and developers do desktop
installs, both for their employer and for themselves.  And they do hack
and customize lots of things.


> They want very few questions and working desktop after the installation.



  This has nothing to do with the possibility of performing any kind of
customization.

  And again, speak for yourself.


> This is probably one of the reason why people who want to do their own
> partition setup are not using desktop-friendly distributions…



  What?  Can you name me one distribution other than Red Hat (which in
fact is not a desktop-friendly distribution) that does not allow one to
"do their own partition setup"?


> You are the first desktop user I have met who insists on having his
> own layout.



  You must have lived in a walled garden then, because everybody I know
who regularly uses some GNU/Linux distribution (for their own use and
often at work) regularly does some level of customization of key
elements of the OS: the kernel, the filesystem, PAM settings, the
initramfs, init, service configuration (both SysV init scripts and
systemd unit files) etc.

  This is in fact one of the major reasons they use Linux, one of them I
remember saying that he switched to Linux when he quickly got bored at
how little his brand new Windows 98 installation allowed him to do.


>>   This is my precisely my point: trumping desktop users' needs (or just
>> freedom of customization and choice)  because of cluster ease-of-use
>> considerations make the distribution *not* universal.  If Devuan is
>
> Of course it is still universal, or does your desktop (whatever you
> use) stop working after a /usr-merge?



  Of course it is no longer universal when I can no longer have a split
/usr filesystem because it makes management of datacenter clusters more
difficult.


> No one removes fvwm, xfce, kde, or the X server.



  Straw man attack: I never said the merge prevents me to install this
or any other software.  I listed the things it prevents me to do several
times, why are you ignoring those points and make up never before stated
claims?


> And as Roger has tried to tell you every freedom of customization and
> choice you want is work that has to be done by someone.



  Don't you tell me!

  Even taking away the possibilities I have to customize installations
"is work that has to be done by someone".


> More choices means you need more testing.



  Err, mostly not.  A distribution can (and in fact they do) have a
default layout that {could be|is} the only officially supported one.

  I am fine with testing and troubleshoot my own customizations and
hacks, I do not ask others to do them for me.  If for no other reason
that they'd take the fun away.  I just want to be able to do them.  And
to have some documentation available, next.  Many critics of systemd
correctly pointed out how little it lets you customize *your* system,
often making it more fragile.  I run into this problem myself, with a
Fedora installation, when systemd was waiting minutes for the swap
partition activation to time out at boot.  This was due to the fact that
I had encrypted it's partition, which was working well after the system
booted.  But the swap.service had nothing at all that could be
customized, it could not even de disabled, it was a thoroughly
internally managed service.  This is one of the reasons I decided to
ditch Fedora and to only use systemd-free distributions.  Now Debian is
going the Red Hat way to prevent you from booting on an unmerged
filesystem.  If this became impossible or too hard and long to undo, I'm
not going back to that distribution.  Would Devuan follow it's example
(which hopefully it will not, looks like is not going to be, not soon
anyway) I'll ditch it too.


> Since Devuan and Debian are build by volunteers they will do what they
> want to do.



  So what?  How many times do I need to re-state that I am not trying to
impose others my way, I just want to be let to choose, to let what has
been possible to do for decades be?


> If no one is interested in keeping the choices they will fade away.
> Will you step forward and work to keep the choices?



  Why are people working to prevent the choices in the first place?


> I don’t know about distribution statistics but in the usr-merge thread
> (debian-devel) it was mentioned that in future distributions are more
> a source to build containers



  Yes, I did notice how much Microsoft have been pushing their: "Hey
look! Windows 10 can even run Linux!"


> and are less and less used directly.



  Well, I am using it directly.  Again, I am not running Devuan on a
datacenter to be managed by a Windows/OSX PC.

  You are confirming that GNU/Linux is been reengineered to just fit use
in a datacenter.

  You indirectly confirm that desktop users are being deliberately left
behind, as they are a niche segment that does not provide a strong
cash-flow to distribution-makers.  Fine.  I'll look for a desktop
Unix-like system elsewhere, designed to be used directly by people, not
only for datacenters.


> And that they have lost users who found Debian not modern enough.



  My blue-jeans too aren't any modern, in fact they are
undistinguishable from those my father wore in his youth.

  "Not modern enough" without further details is CEO and marketing
speech, not technical speech.

  And again, I do get the technical reasons that have datacenter and
cluster sysadmins prefer a merged filesystem, but they do no fit my use
of GNU/Linux as a personal workstation, as a system that is often used
connected to the public Internet and/or to public/unkown/potentialy
hostile networks and must be flexible enough to be rapidly converted to
fit a number of different use cases, unlike a datacenter.


> I don’t know the future but a distribution can’t wait too long to
> choose its direction.



  I know, and in fact I often find myself to have to chose a specific
"direction" for my own installs, because I do not expect distribution
developers to cater to my specific needs.

  But were they to prevent me from doing the kind of customizations that
have been easy to perform since the earliest distributions were out
because "they are old-fashioned", "not enough modern", "unfit to the
datacenter", "unreasonable", "why would one want to do that", well, I'll
jump ship and switch to other hacker-friendly distribution, or I'll
reconsider some BSD distribution, maybe.


>>   Yes, we are.  Debian apparently is no longer going to be.  And again I
>
> Bullshit, that isn’t decided yet.



  The discussion is about going that way.  That's the reason i wrote
"apparently", in the sense of: "according to appearances, initial
evidence, incomplete results, etc.; ostensible rather than actual".


> And for now it would seem that the decision for a forced usr-merge is
> postponed.



  Postponed means:

1. it was proposed;
2. it's still on the table for future implementation.


  I'm speaking out loud now before it's too late and final decisions are
taken that will affect my customizations adversely.



> It can be that they’ll do it like the migration from /usr/doc to
> /usr/share/doc: slowly moving files from / to /usr until / has only
> symlinks.
>
>> was listed the many good reasons the merge is good for the datacenter as
>> an answer to my question: "Why must I be denied the possibility to do
>> otherwise?".
>
> Again, are you doing the work for your setup you care so much?



  Yes, I am.  It's me who designs the filesystem layout when I chose
"Manual" in the install program menu.

  It's me who reconfigures the kernel to fit my specific use-cases.

  It's me who removes and installs software after checking the default
packages that were installed, it's me who hacks App-armor recipies, PAM
settings, iptables chains, security/limits settings, login.defs
settings, udev rules, system services, server configs and so on.  I do
not expect the distribution maintainers to do this for me.  I just want
to keep being free to do it.


>>   So, regardless from what's best to the datacenter, can I be allowed to
>> follow the 40 years long path and keep having a /usr split from /?
>>   Yes && I stay || I leave.
>
> The choice will probably stay as long as the work to keep it won’t be
> too much work for the volunteers.



  And the day the choice will be gone I will surely jump ship.



  Bye,



--
Alessandro Selli <alessandroselli@???>
VOIP SIP: dhatarattha@???
Chiave firma e cifratura PGP/GPG signing and encoding key:
BA651E4050DDFC31E17384BABCE7BD1A1B0DF2AE