5 ms·

""We have already voted with apt: less than 1% of new installs use sysvinit"" The Debian developers voted by setting the apt default. So no.

by yarrel 7y ago

lucb1e 7y ago
I'm sure we can also find a popular package that isn't default. People do install what they like or really care about. I don't particularly like systemd, but I also don't want to bother switching (better one standard that we improve than fragmented and poorly supported competing systems), so it's okay.
- groestl 7y ago
  There's a difference between installing a tool that everyone wants, and replacing the init system of a distro. Honestly, I'm surprised that even 1% replaced the default there.
DiabloD3 7y ago
At work, we use Debian. I already voted by removing systemd and moving to sysvinit during the period where debian-installer did not give me an option. systemd is a constant security hazard, fails standard unix philosophy by doing one thing and one thing well, and is nigh impossible to audit. systemd is a mistake, adopting it was a mistake, and it gives me nothing that either I don't need or want (not cloud/large orchestration stuff; standalone machines), or DO need or want but actual orchestration stuff gives me while also giving me tiny init systems (in some cases, yes, neutered systemd to limit the security implications and shear number of bugs) to bring my stripped down micro-instances up. So, the Debian community voted, and for similar reasons as I voted. We don't want systemd, we never wanted systemd, and we will burn the distro down to get what we want. Redhat tried to use systemd to Microsoft-style Embrace, Extend, Extinguish their way into the entire Linux community, and we want nothing of it.
- nine_k 7y ago
  Have you tried Devuan? (I did not.) It seems to be the closest a Debian derivative without systemd but with everything else intact.
  - DiabloD3 7y ago
    Not directly, but given my choices in configuration in Debian (such as keeping sysvinit over systemd), I basically did, just not in the way you meant.
- mkesper 7y ago
  Alternate init systems (upstart etc.) have been trying to fix the sysvinit mess and failed to gain traction. Then systemd came along with a 'we do a little bit more' approach and all relevant distributions (yes, really) could agree upon using it. I'm very happy the Linux world did something contrary to the Unix wars where everyone was trying to reinvent the wheel until everything collapsed.
  - dredmorbius 7y ago
    Gaining traction often has very little to do with technical superiority, especially when considering long-term fragility or monopolisation of a space. Systemd has been relentlessly pursued by Red Hat. It's won over some technical advocates, but created numerous enemies and critics. Ted T'so is quoted in the LWN piece. Linus himself has been diplomatic but pretty critical, at least in past (I've not seen any specific comments in the past 2-4 years AFAIR). When it first appeared on the scene, systemd was promised as a way of shaving a few seconds off of boots. That's not something I was ever particularly interested in (my systems reboot infrequently, I'm more concerned that they come up clean, reliably, and deterministically, than fast). Meantime, systemd is now taking over my logs, network, device tree, time service, cron, and more. The walk's not matching the talk. The complexity and insecurities remain. And yes, trying to maintain backwards (to sysvinit) or sidewards (to upstart, if that's even still alive, openrc, runit, eudev, procd, nosh, finit, ...) is difficult. It also concedes the initiative (init-iative?) to systemd. Standards wars against well-heeled forces are exceedingly difficult. Even when technology and history are on your side. And now for my talk on the year of Linux on the desktop....
    - Quekid5 7y ago
      > Meantime, systemd is now taking over my logs, network, device tree, time service, cron, and more. This is a misrepresentation. It does have components for all of those things, but AFAIK only journald and udev (I think) are non-optional. You're free to use cron, chrony (NTP), NetworkManager (or whatever), etc.
      - dredmorbius 7y ago
- pjmlp 7y ago
  The myth of UNIX philosophy, that is why Solaris, HP-UX, Aix, DG-UX, SGI ... had so many tools with endless list of options and their own service managers much like in what systemd has followed up on.
  - dredmorbius 7y ago
    The Unix philosophy was always a guideline and goal, not a rigid rule. The usual exceptions to the rule have generally been complexity hubs -- places where a bunch of pieces come together and need to be managed. The shell, editors, email (crossing networking, domain services, local and remote message processing, security, privacy, spam and filtering), compilers, databases, graphics editors, are all examples. But where small and simple tools can be used they prove an excellent idea. systemd took a set of small, simple, and fairly independent concepts and bound them up in both complexity and tight dependencies. The fact that even identifying, splitting out, and replicating functionality outside the system is a problem is a symptom of this fact.
chungy 7y ago
If there was significant desire for something other than systemd, it would be quite a bit higher than 1% being tallied.

""We have already voted with apt: less than 1% of new installs use sysvinit"" The Debian developers voted by setting the apt default. So no.

by yarrel 7y ago

""We have already voted with apt: less than 1% of new installs use sysvinit"" The Debian developers voted by setting the apt default. So no.

lucb1e 7y ago

I'm sure we can also find a popular package that isn't default. People do install what they like or really care about. I don't particularly like systemd, but I also don't want to bother switching (better one standard that we improve than fragmented and poorly supported competing systems), so it's okay.

groestl 7y ago
There's a difference between installing a tool that everyone wants, and replacing the init system of a distro. Honestly, I'm surprised that even 1% replaced the default there.

DiabloD3 7y ago

At work, we use Debian. I already voted by removing systemd and moving to sysvinit during the period where debian-installer did not give me an option. systemd is a constant security hazard, fails standard unix philosophy by doing one thing and one thing well, and is nigh impossible to audit. systemd is a mistake, adopting it was a mistake, and it gives me nothing that either I don't need or want (not cloud/large orchestration stuff; standalone machines), or DO need or want but actual orchestration stuff gives me while also giving me tiny init systems (in some cases, yes, neutered systemd to limit the security implications and shear number of bugs) to bring my stripped down micro-instances up. So, the Debian community voted, and for similar reasons as I voted. We don't want systemd, we never wanted systemd, and we will burn the distro down to get what we want. Redhat tried to use systemd to Microsoft-style Embrace, Extend, Extinguish their way into the entire Linux community, and we want nothing of it.

nine_k 7y ago
Have you tried Devuan? (I did not.) It seems to be the closest a Debian derivative without systemd but with everything else intact.
- DiabloD3 7y ago
  Not directly, but given my choices in configuration in Debian (such as keeping sysvinit over systemd), I basically did, just not in the way you meant.
mkesper 7y ago
Alternate init systems (upstart etc.) have been trying to fix the sysvinit mess and failed to gain traction. Then systemd came along with a 'we do a little bit more' approach and all relevant distributions (yes, really) could agree upon using it. I'm very happy the Linux world did something contrary to the Unix wars where everyone was trying to reinvent the wheel until everything collapsed.
- dredmorbius 7y ago
  Gaining traction often has very little to do with technical superiority, especially when considering long-term fragility or monopolisation of a space. Systemd has been relentlessly pursued by Red Hat. It's won over some technical advocates, but created numerous enemies and critics. Ted T'so is quoted in the LWN piece. Linus himself has been diplomatic but pretty critical, at least in past (I've not seen any specific comments in the past 2-4 years AFAIR). When it first appeared on the scene, systemd was promised as a way of shaving a few seconds off of boots. That's not something I was ever particularly interested in (my systems reboot infrequently, I'm more concerned that they come up clean, reliably, and deterministically, than fast). Meantime, systemd is now taking over my logs, network, device tree, time service, cron, and more. The walk's not matching the talk. The complexity and insecurities remain. And yes, trying to maintain backwards (to sysvinit) or sidewards (to upstart, if that's even still alive, openrc, runit, eudev, procd, nosh, finit, ...) is difficult. It also concedes the initiative (init-iative?) to systemd. Standards wars against well-heeled forces are exceedingly difficult. Even when technology and history are on your side. And now for my talk on the year of Linux on the desktop....
  - Quekid5 7y ago
    > Meantime, systemd is now taking over my logs, network, device tree, time service, cron, and more. This is a misrepresentation. It does have components for all of those things, but AFAIK only journald and udev (I think) are non-optional. You're free to use cron, chrony (NTP), NetworkManager (or whatever), etc.
    - dredmorbius 7y ago
pjmlp 7y ago
The myth of UNIX philosophy, that is why Solaris, HP-UX, Aix, DG-UX, SGI ... had so many tools with endless list of options and their own service managers much like in what systemd has followed up on.
- dredmorbius 7y ago
  The Unix philosophy was always a guideline and goal, not a rigid rule. The usual exceptions to the rule have generally been complexity hubs -- places where a bunch of pieces come together and need to be managed. The shell, editors, email (crossing networking, domain services, local and remote message processing, security, privacy, spam and filtering), compilers, databases, graphics editors, are all examples. But where small and simple tools can be used they prove an excellent idea. systemd took a set of small, simple, and fairly independent concepts and bound them up in both complexity and tight dependencies. The fact that even identifying, splitting out, and replicating functionality outside the system is a problem is a symptom of this fact.

chungy 7y ago

If there was significant desire for something other than systemd, it would be quite a bit higher than 1% being tallied.