19 ms·

I too used FreeBSD for many things between 4x and 10.x, but have made the switch to Ubuntu in the past couple of years. Things I miss: 1) Jails without having

by RNCTX 6y ago

I too used FreeBSD for many things between 4x and 10.x, but have made the switch to Ubuntu in the past couple of years. Things I miss: 1) Jails without having to have a hypervisor system, Linux doesn't really have anything comparably as easy to use particularly with the iocage project managing the jails. 2) ZFS out of the box supported at at lower level than modules and user tools. ZoL has come a ways but Linux's political opposition to it is always going to be a weight around its neck. FWIW I am using a ZFS pool that has never been wiped/recreated since FreeBSD 8. It was created on that OS and has been migrated between Linux/BSD twice over a span of 11 to 12 years. Things I don't miss: 1) The lack of file/directory monitoring libs and third party applications because, honestly, kqueue sucks. 2) While simple SysV init was simple 15 years ago systemd has come a long way. I hate to admit it, but systemd is the better way at this point because I like having my daemons restart themselves without third party monitors. 3) Brain drain is real, it seems the last time I used FreeBSD that port maintainers are running thin and I don't have a suggestion for how to attract more of them. For many years FreeBSD's package/port system was the best in the world (yes, better than the Linux alternatives like apt and rpm too). I think pkgng's rollout was bungled, though, and while it may be better now it seems to have a lingering cost in port maintainers.

xxpor 6y ago
I didn't think there was any political opposition to ZoL other than the fact that its licence is incompatible with the GPL, and outside of a complete clean room rewrite, there's really nothing anyone can do about it. It's not like you're going to convince Oracle to relicence it.
- yjftsjthsd-h 6y ago
  > I didn't think there was any political opposition to ZoL other than the fact that its licence is incompatible with the GPL, and outside of a complete clean room rewrite, there's really nothing anyone can do about it. I suppose it depends on how you mean that, but I probably disagree; when Greg KH can say, "My tolerance for ZFS is pretty non-existant" [0], and Torvalds says, "Don't use ZFS. It's that simple. It was always more of a buzzword than anything else, I feel... [the] benchmarks I've seen do not make ZFS look all that great. And as far as I can tell, it has no real maintenance behind it any more..." [1], we've gone well beyond "they just can't merge it because of licensing" and into "[political] opposition". > It's not like you're going to convince Oracle to relicence it. I've always wondered at this, actually; why doesn't Oracle relicense it? BTRFS, AFAIK, was started at Oracle but basically flopped; why on earth do they not simply bring over the far-superior ZFS codebase to replace it? They can't possibly be trying to protect Solaris the way Sun did; they've killed off >90% of the Solaris team [2] and AFAIK have made exactly zero effort to sell Solaris for years now. Why wouldn't they get ZFS upstreamed into Linux and make it a core feature of OEL (or whatever they were using BTRFS for)? [0] https://marc.info/?l=linux-kernel&m=154714516832389&w=2 https://marc.info/?l=linux-kernel&m=154714516832389&w=2 [1] https://arstechnica.com/gadgets/2020/01/linus-torvalds-zfs-statements-arent-right-heres-the-straight-dope/ https://arstechnica.com/gadgets/2020/01/linus-torvalds-zfs-s... [2] https://fortune.com/2017/09/05/oracle-layoffs-hardware-solaris/ https://fortune.com/2017/09/05/oracle-layoffs-hardware-solar...
  - xxpor 6y ago
    >I've always wondered at this, actually; why doesn't Oracle relicense it? Honestly, I would assume because Oracle's response to anything charitable or nice is "fuck you"
- trasz 6y ago
  If Oracle relicensed it, it would open the door to litigation, since GPL, differently from CDDL, doesn't include a patent grant.
  - xxpor 6y ago
    That's a really good point. I guess they could include a patent grant in addition?
stiray 6y ago
Jails? Hypervisor? :D :D This post is a joke right? Just for anyone else: jails had NEVER had anything to do with hypervisor, it is a kernel level isolation, actually far more complete than cgroups.
- fistikcisahap 6y ago
  I think that's exactly what OP meant when they said they are missing Jails.
  - RNCTX 6y ago
    Yes, exactly. There's no need for a hypervisor to simply isolate one set of services and/or a sandbox for testing, a jail is a simpler solution.
    - stiray 6y ago
      Ouch... diagonal reading kills :) I am sorry :)
JdeBP 6y ago
It's highly misleading to state that one will not miss "simple System V init" from FreeBSD, because it is a falsehood by implication. One of the things that characterizes the BSDs is that they never had AT&T System 5 init, nor the van Smoorenburg clone thereof for Minix+Linux. It's one of the things that the BSD side of the universe resisted from the AT&T side of the universe back in the 1980s. FreeBSD has BSD init, a rather different beast to AT&T init, and Mewburn rc, which is very different to van Smoorenburg rc (although the post-2014 revisions on Debian make van Smoorenburg rc come closer) on Linux. And it's equally misleading to talk about liking your dæmons restarting themselves, because that's untrue too. systemd is the very third party monitor, externally monitoring and restarting the dæmon programs so that they do not have to restart themselves, that you claim to want to do without.
- RNCTX 6y ago
  Sorry, I don't care about branding semantics. The difference between inittab and the scripts controlling the startup routines are minimal. A person in 2005 could grasp the difference between BSD and Linux init scripts in about an hour. As for hating systemd, that battle is lost, my friend. There's really no point in even engaging any discussion about it beyond this paragraph. No one is looking to write shell scripts for init in 2020 when there are systemd service files that are simple configs. It's provided by the distro, that's why it's adoptable. No one in the right mind trusts open source project devs to maintain compatibility in even minor version variations, that's why there's a booming container business. Tools that the distro provides work, tools from third parties are on borrowed time until they break, so yes systemd is the only viable up-monitoring solution.
  - MintelIE 6y ago
    Many people have rejected the systemd cult of bloat and scripts are easy to read, write, understand, and maintain. Most consist of generic boilerplate anyway. Nobody wants to replace a simple init system with something with 1.5 million lines of code like systemd. For crying out loud, it’s 2020.
    - 6y ago
butterisgood 6y ago
Kqueues are great!
- JdeBP 6y ago
  kevent() is indeed the MsgWaitForMultipleObjects() of the Unix world, a very useful mechanism, and indeed the very file directory monitoring library function that RNCTX wrongly claims to be lacking. But from experience of using it extensively in my softwares I can report that it has some warts on FreeBSD. A very quick précis: * It has problems if one wants to combine NOTE_TRACK, NOTE_FORK, and NOTE_EXIT in EVFILT_PROC, because it returns one event for both a forked child and an exited child, which both try to store different things in the single "data" field of the event. * Quite a few of the pollable device drivers in FreeBSD are not kevent() enabled. So it mysteriously fails to work with a whole bunch of I/O devices, where (confusingly for diagnosis) polling does. kevent() is not universally usable. * One might think, from reading the various operating system's manual pages, that only NetBSD and MacOS have EVFILT_FS. In fact, FreeBSD kevent() also has EVFILT_FS. It's just entirely undocumented.
- bnastic 6y ago
  Yes, kqueue is very good, just not for the purpose in the OP's context - filesystem monitoring.
Grimm665 6y ago
I admit to not being as familiar with Jails as I'd like, but systemd-nspawn seems to cover much of this use-case on the Linux side. I've used it in a few different scenarios and found it very easy and intuitive to set up. Easier and simpler than VZ for kernel-based VMs for sure.

by RNCTX 6y ago

xxpor 6y ago

I didn't think there was any political opposition to ZoL other than the fact that its licence is incompatible with the GPL, and outside of a complete clean room rewrite, there's really nothing anyone can do about it. It's not like you're going to convince Oracle to relicence it.

yjftsjthsd-h 6y ago
> I didn't think there was any political opposition to ZoL other than the fact that its licence is incompatible with the GPL, and outside of a complete clean room rewrite, there's really nothing anyone can do about it. I suppose it depends on how you mean that, but I probably disagree; when Greg KH can say, "My tolerance for ZFS is pretty non-existant" [0], and Torvalds says, "Don't use ZFS. It's that simple. It was always more of a buzzword than anything else, I feel... [the] benchmarks I've seen do not make ZFS look all that great. And as far as I can tell, it has no real maintenance behind it any more..." [1], we've gone well beyond "they just can't merge it because of licensing" and into "[political] opposition". > It's not like you're going to convince Oracle to relicence it. I've always wondered at this, actually; why doesn't Oracle relicense it? BTRFS, AFAIK, was started at Oracle but basically flopped; why on earth do they not simply bring over the far-superior ZFS codebase to replace it? They can't possibly be trying to protect Solaris the way Sun did; they've killed off >90% of the Solaris team [2] and AFAIK have made exactly zero effort to sell Solaris for years now. Why wouldn't they get ZFS upstreamed into Linux and make it a core feature of OEL (or whatever they were using BTRFS for)? [0] https://marc.info/?l=linux-kernel&m=154714516832389&w=2 https://marc.info/?l=linux-kernel&m=154714516832389&w=2 [1] https://arstechnica.com/gadgets/2020/01/linus-torvalds-zfs-statements-arent-right-heres-the-straight-dope/ https://arstechnica.com/gadgets/2020/01/linus-torvalds-zfs-s... [2] https://fortune.com/2017/09/05/oracle-layoffs-hardware-solaris/ https://fortune.com/2017/09/05/oracle-layoffs-hardware-solar...
- xxpor 6y ago
  >I've always wondered at this, actually; why doesn't Oracle relicense it? Honestly, I would assume because Oracle's response to anything charitable or nice is "fuck you"
trasz 6y ago
If Oracle relicensed it, it would open the door to litigation, since GPL, differently from CDDL, doesn't include a patent grant.
- xxpor 6y ago
  That's a really good point. I guess they could include a patent grant in addition?

stiray 6y ago

Jails? Hypervisor? :D :D This post is a joke right? Just for anyone else: jails had NEVER had anything to do with hypervisor, it is a kernel level isolation, actually far more complete than cgroups.

fistikcisahap 6y ago
I think that's exactly what OP meant when they said they are missing Jails.
- RNCTX 6y ago
  Yes, exactly. There's no need for a hypervisor to simply isolate one set of services and/or a sandbox for testing, a jail is a simpler solution.
  - stiray 6y ago
    Ouch... diagonal reading kills :) I am sorry :)

JdeBP 6y ago

It's highly misleading to state that one will not miss "simple System V init" from FreeBSD, because it is a falsehood by implication. One of the things that characterizes the BSDs is that they never had AT&T System 5 init, nor the van Smoorenburg clone thereof for Minix+Linux. It's one of the things that the BSD side of the universe resisted from the AT&T side of the universe back in the 1980s. FreeBSD has BSD init, a rather different beast to AT&T init, and Mewburn rc, which is very different to van Smoorenburg rc (although the post-2014 revisions on Debian make van Smoorenburg rc come closer) on Linux. And it's equally misleading to talk about liking your dæmons restarting themselves, because that's untrue too. systemd is the very third party monitor, externally monitoring and restarting the dæmon programs so that they do not have to restart themselves, that you claim to want to do without.

RNCTX 6y ago
Sorry, I don't care about branding semantics. The difference between inittab and the scripts controlling the startup routines are minimal. A person in 2005 could grasp the difference between BSD and Linux init scripts in about an hour. As for hating systemd, that battle is lost, my friend. There's really no point in even engaging any discussion about it beyond this paragraph. No one is looking to write shell scripts for init in 2020 when there are systemd service files that are simple configs. It's provided by the distro, that's why it's adoptable. No one in the right mind trusts open source project devs to maintain compatibility in even minor version variations, that's why there's a booming container business. Tools that the distro provides work, tools from third parties are on borrowed time until they break, so yes systemd is the only viable up-monitoring solution.
- MintelIE 6y ago
  Many people have rejected the systemd cult of bloat and scripts are easy to read, write, understand, and maintain. Most consist of generic boilerplate anyway. Nobody wants to replace a simple init system with something with 1.5 million lines of code like systemd. For crying out loud, it’s 2020.
  - 6y ago

butterisgood 6y ago

Kqueues are great!

JdeBP 6y ago
kevent() is indeed the MsgWaitForMultipleObjects() of the Unix world, a very useful mechanism, and indeed the very file directory monitoring library function that RNCTX wrongly claims to be lacking. But from experience of using it extensively in my softwares I can report that it has some warts on FreeBSD. A very quick précis: * It has problems if one wants to combine NOTE_TRACK, NOTE_FORK, and NOTE_EXIT in EVFILT_PROC, because it returns one event for both a forked child and an exited child, which both try to store different things in the single "data" field of the event. * Quite a few of the pollable device drivers in FreeBSD are not kevent() enabled. So it mysteriously fails to work with a whole bunch of I/O devices, where (confusingly for diagnosis) polling does. kevent() is not universally usable. * One might think, from reading the various operating system's manual pages, that only NetBSD and MacOS have EVFILT_FS. In fact, FreeBSD kevent() also has EVFILT_FS. It's just entirely undocumented.
bnastic 6y ago
Yes, kqueue is very good, just not for the purpose in the OP's context - filesystem monitoring.

Grimm665 6y ago

I admit to not being as familiar with Jails as I'd like, but systemd-nspawn seems to cover much of this use-case on the Linux side. I've used it in a few different scenarios and found it very easy and intuitive to set up. Easier and simpler than VZ for kernel-based VMs for sure.