Yea, I saw that earlier this morning, and that sounds somewhat promising. But I just went to http://developer.apple.com , searched for ZFS, and got 0 results found.
Has anyone found the ZFS kext and installed it? Any success?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
No ZFS Support in Leopard?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Yea, I saw that earlier this morning, and that sounds somewhat promising. But I just went to http://developer.apple.com , searched for ZFS, and got 0 results found.
Has anyone found the ZFS kext and installed it? Any success?
Wow. This whole year has been pretty dissapointing as far as mac COMPUTING goes. I wonder if it has anything to do with Apple over extending themselves by programming for WINDOWS!? (ie Safari and iPhone compatibility)
Where is iLife BTW?
Where is iLife BTW?
It's on the phone
It's probably been pulled for inclusion on the phone thingy
Great
I'd prefer if they spend their times on this, than on expanding Photo Booth making it able to look like a shark in iChat. Very disappointing.
It's probably been pulled for inclusion on the phone thingy
Yea... And where is the Mac MiniTower for ~$900? And where is the Apple Home Server? Ha! Steve Jobs can't be serious about "throwing" a USB hard drive onto an AirPort Extreme Base Station as a solution to the home network storage/archive/backup need. What a @#$%ing joke
sweet!
from memory, previous attempts at running ZFS utils in leopard devs returned a "not implemented" message.
seems to me that ZFS is somewhat there, maybe they're working on the final touches? maybe it will be "enabeld" via a 10.5.x update?
Because this it rather off topic in a thread on ZFS I've taken it to a separate thread: https://forums.macrumors.com/threads/313411/
Yes, that's exactly what's happened. Apple [=SJ] is delivering a message to everyone: obey your NDA... and don't steal our thunder.
I'm not techie enough to analyze the details of the code. I AM user enough to opine that the "minor tweaks" announced for the new OSX amount to nothing more than a modest version upgrade, comparable to each of the moves from 10.3 to 10.4 etc., and giving me nothing more than incremental improvements.
I thought the new software was to be a bigger step forward. I took Jobs' and the company's earlier remarks to indicate this new version of OSX would significantly improve my user experience. Did i misunderstand? Isn't this, at best, the most over-promised new Apple product in a decade?
If not, can readers point to ways in which this announcement indicates greater improvements than we've seen with each of the prior versions? I want a new, far-superior product. I'm having a real hard time believing that's what I'm going to get.
Nevermind the fact current Solaris betas ARE able to boot ZFS, ZFS doesn't have to support boot volumes - no filesystem does, as that's not a function of the filesystem. The boot loader needs to understand how to boot a filesystem. The reason it took so long to get ZFS to boot on Solaris/x86 was due to GRUB not supporting it. It's still not bootable on Sparc because OpenFirmware doesn't support it yet. To get ZFS to boot, Apple would just have to update the OS X loader.
ZFS will be a real kernel-mode filesystem in FreeBSD 7 (actually, it's currently implemented in the dev branch). Linux support will have to be done via something like FUSE, unless Sun relicenses ZFS to be GPL-compatible.
You really don't have to do that now, you could make one big / partition if you wanted (swap, obviously has to be separate).
I have difficulty with the "Jobs pulled it because Sun pre-announced it" stuff too.
I think the most likely reason for it not being an officially announced part of Leopard is that the XNU port just isn't production quality yet.
File systems are complex and important components. A single bug can cause massive irrecoverable file loss. ZFS may or may not be the future of Mac OS X, but you can be sure Apple isn't going to just throw it out and hope it works.
Given Apple's known problems with resources at the moment, with Leopard delayed because of iPhone etc, it's not exactly a surprise that ZFS isn't ready.
I think the most likely reason for it not being an officially announced part of Leopard is that the XNU port just isn't production quality yet.
File systems are complex and important components. A single bug can cause massive irrecoverable file loss. ZFS may or may not be the future of Mac OS X, but you can be sure Apple isn't going to just throw it out and hope it works.
Given Apple's known problems with resources at the moment, with Leopard delayed because of iPhone etc, it's not exactly a surprise that ZFS isn't ready.
What applications would a new filesystem break? Only the low-level applications that mess with the filesystem. Most applications don't touch the filesystem; they touch the VFS layer, which abstracts the filesystem. The only issue is case sensitivity, but another person in this forum claims that most applications work fine with case-sensitive HFS. While I use case-sensitive HFS for my /Users partition, I keep / with the default; I can't say whether my applications would break.
The only other problem would be resource forks, but there are a number of ways Apple can handle that.
Aside from file case, applications are ignorant of the details of a filesystem!
Well in FreeBSD you could always put everything in one big partition. It just hasn't been advisable and a lot of those reasons are still valid with ZFS. To name a few: a stuck server process causing a log file running wild in /var/log would not fill up all the space in the more important file systems, if you keep them separate. Also, if you have separate partitions you can unmount them separately or even mount certain partitions in read-only mode, which can be handy for security reasons.
We just need to wait and see what the developers copy has in it. It shouldn't be more than a couple days before people announce what's in the "feature complete" copy.
Well, applications should be ignorant of the details of their filesystem. But as the case thing shows, this is not always so. There could be other glitches as well, but you won't know until you do a thorough beta test.
All I was saying in my original post was that I couldn't see Apple throw ZFS in the mix as the default filesystem at the very last minute before the release. There just would be too many things that could go wrong. Beta testing is an important part of the QA process and I can't imagine that Apple would skip it for something as important as a filesystem. I was saying this in response of people commenting that we might still get ZFS as the default filesystem. I really don't think so. Perhaps as an option, and I sincerely hope so. But if it doesn't start to fully work in the developer builds any day now, I couldn't see it happening. These things really need a few months of testing before they can be put in a final build.
I know it has been in some builds up till now but it was always broken.
Yeah, I was running ZFS in FreeBSD-CURRENT a few weeks ago, I thought it was great (snapshots rule!
) and Pawel's porting effort has been amazing. The current problems for ZFS on FreeBSD is that, at least on the 32bit version of FreeBSD, it needs a very large amount of memory to run effectively; I'm sure it'll be fixed when FreeBSD 7 is released toward the end of the year (probably about the same time as Leopard interestingly).If ZFS does not make it into the initial release of Leopard, i'd suggest that for those who it was important install FreeBSD 7 (when its out) and give it a whirl; no harm in practising before ZFS comes to Mac OSX which I'm certain it will at some point.
I run my entire drive (only 1 partition) as case-sensitive HFS. The ONLY things I've every come across that have problems is adobe apps (at least CS2 and earlier, i haven't tried CS3), and the image capture plugin for epson scanners (included with OS X).
I think we will see ZFS in Leopard, when who knows. I think Apple is saying that it will be omitted based on the fact that Sun announced it before Apple could. So, in order for Apple to announce it, they will make it appear that it will not be in it and then later announce that they have implemented it into 10.5. This will help bring back the "thunder" to Apple that everyone says was lost because of Sun's announcement.
For something as important as a filesystem, you don't want to change it out from under everybody at once. Apple's been smart about this before, so I expect ZFS will be optional for quite a while before it becomes default. I wouldn't expect it to be default any time before 10.6, after everybody has a chance to see how well it interacts with the Mac's application base.
People who think apple would completely change plans because someone at another company made comments about a product are completely absurd.
Do you really think any company would scrap months (maybe even years) of work because of something someone said?
That's not the way business works. Once plans are laid down, dollars committed, and people allocated, only major obstacles will prevent those plans/dollars/people from following through.
And that dude's comments, saying "ZFS is not happening", that doesn't mean it's not going in leopard. It could just mean zfs is not the most exciting thing going on, as in, hey, it's not "happening" as in, hey, what's happening... ?
Do you really think any company would scrap months (maybe even years) of work because of something someone said?
That's not the way business works. Once plans are laid down, dollars committed, and people allocated, only major obstacles will prevent those plans/dollars/people from following through.
And that dude's comments, saying "ZFS is not happening", that doesn't mean it's not going in leopard. It could just mean zfs is not the most exciting thing going on, as in, hey, it's not "happening" as in, hey, what's happening... ?
On Case Sensitivity
I'm not arguing that new filesystem code needs thorough testing. I whole-heartedly agree that filesystems should be extremely thoroughly tested, because of the potential risks involved.
But no modern application should deal directly with the filesystem, aside from a low-level disk manipulation utility. Apple's documentation at http://developer.apple.com/technotes/tn/tn2037.html tells me that file operations pass through the BSD subsystem, even with legacy code. Hence, even older applications go through the abstract VFS layer.
The only way applications can break, therefore, is if the filesystem does not support a feature expected by the application, and the VFS cannot emulate it. For example, the VFS has to map resource forks into something recognizable to ZFS (personally, I think resource forks should disappear). However, the programs still only see an abstracted representation of these features, and are unaffected by the actual filesystem implementation.
Case sensitivity is a special problem, because it is a "feature" of the filesystem that conflicts with another "feature": case insensitivity. Furthermore, due to its nature, the VFS must also choose between case sensitivity and insensitivity. With regular HFS and HFS+, the VFS could always be case sensitive (a feature remaining from the VFS's BSD origins), while secure in the knowledge that the low-level filesystem driver would be insensitive.
Thus, provided Apple can modify the VFS-to-filesystem interface to map all high-level features of the VFS onto ZFS filesystems, there will be no problems with applications. Case sensitivity is only a problem because it is mutually exclusive with case insensitivity.
Of course, the VFS could mimic case insensitivity, if desired, by casing all filenames into upper- or lower-case. However, that is a bad idea. Despite what long-time proponents of case-insensitive filenames say, there is some utility in recognizing the difference between Complex.h and complex.h.
Here is a guy who talks more about his issues with case sensitivity:
I'm not arguing that new filesystem code needs thorough testing. I whole-heartedly agree that filesystems should be extremely thoroughly tested, because of the potential risks involved.
But no modern application should deal directly with the filesystem, aside from a low-level disk manipulation utility. Apple's documentation at http://developer.apple.com/technotes/tn/tn2037.html tells me that file operations pass through the BSD subsystem, even with legacy code. Hence, even older applications go through the abstract VFS layer.
The only way applications can break, therefore, is if the filesystem does not support a feature expected by the application, and the VFS cannot emulate it. For example, the VFS has to map resource forks into something recognizable to ZFS (personally, I think resource forks should disappear). However, the programs still only see an abstracted representation of these features, and are unaffected by the actual filesystem implementation.
Case sensitivity is a special problem, because it is a "feature" of the filesystem that conflicts with another "feature": case insensitivity. Furthermore, due to its nature, the VFS must also choose between case sensitivity and insensitivity. With regular HFS and HFS+, the VFS could always be case sensitive (a feature remaining from the VFS's BSD origins), while secure in the knowledge that the low-level filesystem driver would be insensitive.
Thus, provided Apple can modify the VFS-to-filesystem interface to map all high-level features of the VFS onto ZFS filesystems, there will be no problems with applications. Case sensitivity is only a problem because it is mutually exclusive with case insensitivity.
Of course, the VFS could mimic case insensitivity, if desired, by casing all filenames into upper- or lower-case. However, that is a bad idea. Despite what long-time proponents of case-insensitive filenames say, there is some utility in recognizing the difference between Complex.h and complex.h.
Here is a guy who talks more about his issues with case sensitivity:
Migration
I get the feeling that we'll probably see Apple trying to migrate everyone to Case-Sensitive Journaled HFS+ before we go to a case-sensitive non-HFS filesystem. This would give developers the opportunity to revamp their code and also would make it such that a transition to ZFS would not be so jarring.
I get the feeling that we'll probably see Apple trying to migrate everyone to Case-Sensitive Journaled HFS+ before we go to a case-sensitive non-HFS filesystem. This would give developers the opportunity to revamp their code and also would make it such that a transition to ZFS would not be so jarring.
I believe the case sensitivity issue is moot. From what I can recall, case insensitivity support was requested, fasttracked, and committed to ZFS source over a month ago.