Linus Torvalds Doesn't Recommend Using ZFS On Linux

Status
Not open for further replies.
Working primary for a hosting company I can say that the usual Linux FS are very limited in their stability. On own projects or earlier projects and some of friends mostly used ZFS - and with ZFS there were never any problem.

You want to create a bigger raid with SSD Cache on Linux? Or any other Cache for it? It's risky and make only problems, more than you can earn on performance.
Linux may have some benefits in some points, but finally if you want a very reliable system my experience is on FreeBSD much better! And if you use good and the right Hardware you should never have problems with ZFS - it's just working. And I think I'm not alone with this opinion.
In my opinion it's the philosophy behind the development. (Lot of you already wrote about it in this thread^^)

That's the reason I'm moving away from Linux to FreeBSD based. And lot of friends have done this before.
 
The monstrous, the senseless and merciless GRUB for Zfs on Linix...
When install GRUB involuntarily to grudge that the history didn’t go the other route:
"If 386BSD had been available when I started on Linux, Linux would probably never had happened" © Linus Torvalds 1993 :)
 
teo said:
Linus Torvalds has reasons to say clearly what ZFS is, the reasons are because he doesn't see the file system as such a relevant technology and has no real maintenance. Considering Oracle's litigious, nature given the claim of the oracle interface ( Java ) I don't think it's a real license and performance gain either.
Click to expand...

Assuming Linus Torvalds' misgivings concerning Oracle are valid, could this be a problem for FreeBSD using ZFS, too?
Why would Linux have a problem and not FreeBSD?
 
weberjn said:
Assuming Linus Torvalds' misgivings concerning Oracle are valid, could this be a problem for FreeBSD using ZFS, too?
Why would Linux have a problem and not FreeBSD?
Click to expand...
Because Torvalds has no idea what he's talking about?

His license concerns are about including something not compatible with his GPLv-whatever Linux kernel. That's his problem, not any of the BSDs. He uses that stupid licensing system GPL, so let him deal with it. He also seems to be totally confused about openZFS and ZFS from Oracle. He's the owner of Linux, he's not a file system developer but unfortunately any crap spewed from his mouth is treated as gospel by the Linux believers and this is where all the confusion seems to be. He could solve the problem, immediately, by re-licensing the kernel code to BSD-style. ;)

Anyway, Linux has about 100 different file systems to choose from (and growing), so why should he be worried about ZFS! (Ok the number is probably a bit of an exaggeration...)

ZFS, which is now under control of Linux-style developers will no doubt head down the path of bloat and sloth as they steadily add more and more "features" to it, just like they do to windowing compositors, kernels, init systems, you name it.
 
Hakaba said:
My interest : Is ZFS still considered as an important feature in FreeBSD ?
(The answer is yes with the current informations)

The ideologic war in Opensource licence world is not my cup of tea.

But it is for some people, so if systemd is a very good evolution, why there is no «BSD licenced systemd like» initiative ?
Click to expand...
Not for lack of trying I think.
I watched Benno Rice (previous FreeBSD core member) pour scorn on FreeBSD in his talk (available on youtube) because they (whoever 'they' are) would not countenance FreeBSD adopting something like systemd. He seems to love the concept.

Hopefully his pleas for systemd on FreeBSD have fallen on deaf ears. Apple has launchd, which seems to be the basis for Poettering's systemd and I don't see a lightning fast or particularly stable OS because of it!
 
Hakaba said:
I am not an expert but I like the idea that maunchd or systemd have to be the process with ID 2.
(https://forums.freebsd.org/threads/systemd-vs-launchd.44973/).
Click to expand...

If you're talking about restarting failed programs, why? There are already tools that perform that if you wish. Perp and daemontools, for example.

launchd is a mimic/'enhancement' of SMC from Solaris. SystemD is a mimic of launchd. They all attempt to consolidate various tools into one huge monolith for no other reason, it seems, than they can. It doesn't benefit anyone other than the developers as they scope creep into oblivion. So you might like some of the features of launchd but I'll bet the house's money you'll see scope creep just like systemD if someone decides to use it as a replacement for rc.

When did systemD start? 10 years ago? It's still growing and will not stop until it has all the userland utils (essentially replacing GNU) and boot control.

But that's the Linux way: scope creep and bloat. I hope the same doesn't happen to ZFS.
 
Phishfry said:
Because the OpenZFS license is CDDL. It is compatible (legally similar) with the BSD 2Clause license.
The CDDL license is not compatible with GPL2/3.

A discussion on combining CDDL and GPL code [LWN.net]

lwn.net lwn.net
Click to expand...
And not only that, it seems Sun deliberately designed it that way, taking into consideration the viral nature of the GPL.

The obvious solution is for Torvalds to re-license under BSD. Simple... ;) But then he'd lose control and that's not in his nature. :eek:
 
mark_j said:
If you're talking about restarting failed programs, why?
Click to expand...
No, no... I am talking about the idea : don't touch the id 1 for supervision or other tasks. Let the choice about what init (id=1) launch. If you want a systemd like features, let init (id=1) launch it (and only it so the id of this tools will be 2).
I am sure that the time of this fork call is not mesurable. But the design is more comprehensive and let a lot more liberty (don't use systemd like, use it for all, use it for specific deamons...)
But this tools exists, that stop the need of systemd.
 
Hakaba said:
No, no... I am talking about the idea : don't touch the id 1 for supervision or other tasks. Let the choice about what init (id=1) launch. If you want a systemd like features, let init (id=1) launch it (and only it so the id of this tools will be 2).
I am sure that the time of this fork call is not mesurable. But the design is more comprehensive and let a lot more liberty (don't use systemd like, use it for all, use it for specific deamons...)
But this tools exists, that stop the need of systemd.
Click to expand...

Precisely. There are tools in ports that can accomplish this. Heck, I've got a shell script that's triggered by cron to watch for a daemon and if it doesn't exist it sends me an email. I think sysadmins have been doing this stuff for decades. I personally don't need some monolithic suite of programs with binary files and sets of commands to accomplish a few things I can do in a shell.

Benno Rice, a Freebsd systemd fanboy gives you an idea why these projects are scope creep in the making:
View: https://youtu.be/o_AIw9bGogo?t=2481
 
"Unix is dead"

Why is that guy a FreeBSD developer, he should jump ship to Linux or Apple since he seems to love them so much. It's like he's only in it for the drama of hammering on the philosophy of the FreeBSD community.

Anyway that talk was super depressing for me, what a millennial dick.
 
Since this thread has diverged from the original topic, which is out of our hands anyway, and has now arrived at the plunger level, are there objections in closing it up in the comming days?
 
mark_j said:
He uses that stupid licensing system GPL, so let him deal with it.
[...]
He's the owner of Linux [...] He could solve the problem, immediately, by re-licensing the kernel code to BSD-style.
Click to expand...
mark_j said:
The obvious solution is for Torvalds to re-license under BSD. Simple... ;) But then he'd lose control and that's not in his nature. :eek:
Click to expand...
I don't think it's that simple. There are far too many contributors to Linux and you would need the agreement of all of those contributors in order to relicence it. I don't think it's a case of any kind of ideological adherence to GPL, it's more so that at this point there is no real choice.

Crivens, you may fire when ready...
 
CraigHB said:
"Unix is dead"
Click to expand...
Bruno Ruce is right. It isn't depression, Its real outlook for his side,
How sing one singer
The plastic world won
the dummy turned out to be stronger
Click to expand...
Died HP-UX, Solaris and other big rock monsters.:(
I think that and IBM AIX will died in the near future, inside IBM - "There Can Be Only One" - if IBM want to develop AIX - didn't spend $34 billion on the purchase RedHat.
Only the battered and scattered partisans remained in an enemy environment - BSD systems and Illumos :)
Some are no more, and distant... others ©
 
mark_j said:
Apple has launchd, which seems to be the basis for Poettering's systemd and I don't see a lightning fast or particularly stable OS because of it!
Click to expand...

There is absolutely nothing wrong with event driven boot managers, quite on the contrary. You shouldn't base your opinion about a technology on amateur
implementation[1] (systemd) and in a purpose built desktop only solution (launchd). Have a look at 'Solaris SMF' and I heard 'Amoeba' also have very interesting ideas in this regards.

Vadim_Mkk said:
I think that and IBM AIX will died in the near future, inside IBM - "There Can Be Only One" - if IBM want to develop AIX - didn't spend $34 billion on the purchase RedHat.
Click to expand...

There are some rumors of IBM willing to get rid of AIX since years but buying RedHat will not help with it at all, because the typical AIX consumer (the POWER heavy consumers, aka true mission critical guys) would never run Linux on their environments.

[1] just remembering RedHat make a living from support.
 
Status
Not open for further replies.
Back
Top