Why not use ext3, ext4 or UFS?

[TechieJustin]

Linus Torvalds says the Apple HFS+ is crap. I can't quite grasp why he said this, but alot of people believe it to be true. In my opinion the Journalized HFS+, is working just fine. It isn't any worse than other proprietary filesystems such as NTFS.
But my question is, instead of putting all that work into making HFS+, why not use an already established filesystem?
Back in my computer science undergrad days my Professor always said the most efficient way to write a program was not to do so in the first place; if somebody else already did the same thing. Why reinvent the wheel?
Are there certain limitations to the ext3, 4 or UFS that I don't comprehend?
I tried reading up on all of them and I felt like a freshman all over again.
It seems Apple has something against UFS since it was removed as an option from 10.5 on.

 

[MisterMe]

...
But my question is, instead of putting all that work into making HFS+, why not use an already established filesystem?
...

OK. HFS+ has been around since MacOS 8.1. It now supports more UNIX installations than any other filesystem on Earth. By your own admission, it works well. What can be any more established than that?

 

[TechieJustin]

OK. HFS+ has been around since MacOS 8.1. It now supports more UNIX installations than any other filesystem on Earth. By your own admission, it works well. What can be any more established than that?

But it wasn't really designed for a unix type environment. Wouldn't modifying the BSD kernel to use HFS+ be more trouble than just using UFS or ext*?

 

[robbieduncan]

But it wasn't really designed for a unix type environment. Wouldn't modifying the BSD kernel to use HFS+ be more trouble than just using UFS or ext*?

It might or might not be. Apple didn't have to. OSX doesn't use the BSD kernel: it uses the Mach micro-kernel with some BSD userland.

 

[maflynn]

There's really no real reason why apple isn't using ext4. Perhaps because of the delayed allocation mechanism and the risk of data loss/corruption has them a bit spooked. Early adopters ran into issues with this but it seems the problems have been ironed out.

ZFS is clearly the superior option but apple has already stated that ZFS is persona non grata

What I really hope is that they don't try to build one from scratch. Not that they can't but the time and resources required will mean that we'll not see a new FS for quite some time.

 

[robbieduncan]

I'm kind of intrigued now. People seem to want to replace HFS+. But why? What improvements would you expect to see?

 

[maflynn]

I'm kind of intrigued now. People seem to want to replace HFS+. But why? What improvements would you expect to see?

Take ZFS (I pulled this from wikipedia) and its very incomplete list.
better blocking, variable block sizes up to 128k, HFS+ 512byte blocks
copy-on writing ability, i.e., duplicating the writes, ensures data integrity
End-to-end checksumming, using a kind of "Data Integrity Field", allowing data corruption detection (and recovery if you have redundancy in the pool).
Transparent filesystem compression. Supports LZJB and gzip.

Newer filesystems are better designed to handle the larger hard drives of today and tomorrow then the older FS could. They have better redundancy built in to ensure data integrity. Better algorithms to minimize to a large extent any need to defragment your HD.

 

[gnasher729]

Linus Torvalds says the Apple HFS+ is crap. I can't quite grasp why he said this, but alot of people believe it to be true. In my opinion the Journalized HFS+, is working just fine. It isn't any worse than other proprietary filesystems such as NTFS.
But my question is, instead of putting all that work into making HFS+, why not use an already established filesystem?
Back in my computer science undergrad days my Professor always said the most efficient way to write a program was not to do so in the first place; if somebody else already did the same thing. Why reinvent the wheel?
Are there certain limitations to the ext3, 4 or UFS that I don't comprehend?
I tried reading up on all of them and I felt like a freshman all over again.
It seems Apple has something against UFS since it was removed as an option from 10.5 on.

Linus Torvalds says Apple HFS+ is crap, Steve Jobs says that Linus Torvalds talks crap. In both cases, just because they say so doesn't mean it's true.

And HFS+ _is_ an established file system. It has been in use for fifteen or more years. Now if Apple started writing HFS+ today, you would have a point, but it's been in actual use for ages and ages and ages. So you should really ask, why did Hans Reiser write his file system? Didn't do him much good, he'll spend all or most of his life in jail. Why was Sun writing ZFS? Didn't do the company much good, they've been sold now. When you ask "Why not use an already established file system": Please do a reality check. HFS+ has many times more installations than ext2, ext3, ext4 and UFS combined.

UFS has been causing trouble for everyone using it. If you reported problems with the code that I write, and you told me you are using UFS, the answer would be: "UFS file systems are not supported; use HFS+ instead". UFS runs slower. UFS causes trouble. UFS has no redeeming features compared to HFS+. It only causes costs with no benefits.

There's really no real reason why apple isn't using ext4.

Except the fact that 100% of all MacOS X applications and 100% of all iPhone applications have been tested with HFS+, and approximately 0% have been tested with ext4.

 

[maflynn]

Except the fact that 100% of all MacOS X applications and 100% of all iPhone applications have been tested with HFS+, and approximately 0% have been tested with ext4.

Well, yeah that's a given. We're not talking about testing but technically speaking, there's nothing holding back OSX from using a different file system other then HFS+

Any change in a file system will require a significant amount of testing and possibly updates to any given application. The issue is that while HFS+ "works" its an old FS that's getting harder to keep up with the demands of modern drives and modern OS/Application needs. Just look at how apple enabled compression in HFS+ with SL. Its a kludge, yeah it works but its tacked on in a way that could possibly give people headaches down the road when apple tries to bolt on another update to this aging file system.

 

[Bill Gates]

Ext4 is just a stop-gap measure until BtrFS is stabilized. The developers themselves have stated this. BtrFS is probably the most-likely candidate for an HFS+ replacement.

 

[maflynn]

Any ETA on btrfs? I looked at their wiki and failed to notice any release dates?

 

[TechieJustin]

Perhaps I misspoke when I said an "established" file system. What i meant was an "existing" filesystem.

Regarding ext3/4 I have some questions. Are they "durable?" Can they withstand the rigors or users shutting down their machines improperly on a semi regular basis without blowing up the filesystem? I think every single one of us has run their battery completely dead and have their Macbook shut down. I've done it pretty often lately. It seems filesystems like NTFS, FAT*, HFS+ can withstand that kind of abuse.
Its been said HFS has been around forever, and then they just added journalization. That seems like a band-aid solution to me. Would Microsoft be correct in adding journalization to FAT64 and calling it FAT64+? Or exFAT+?
I think its time to retire what can be considered an outdated filesystem.

Abandoning HFS+ doesn't necessarily admit there's something wrong with it, just that something better came along. Eventually we're going to reach a point where HFS+ won't work. We ran into that with FAT16 in the mid 1990's. Then again with FAT32 (files bigger than 4GB?).

As for what advantages would be gained in switching to a newer and more robust filesystem, ease of development, speed, reliability, and compatibility.
as for that last bit, I'm still waiting for an exFAT driver for SL... formatting jumpdrives to NTFS is cumbersome.

Apple shouldn't have to worry about development costs as there are already filesystems out there that will do the job and then some.

 

[maflynn]

ext3 and ext4 have been and continue to be used by many linux distros including ubuntu and fedora. Redhat has been using ext3 for many years so I'd say its "durable"

 

[TechieJustin]

ext3 and ext4 have been and continue to be used by many linux distros including ubuntu and fedora. Redhat has been using ext3 for many years so I'd say its "durable"

It works fine for people who use Linux - which tend to be more knowledgeable.
Would ext hold up to the rigors of the Poli Sci major fo example? I have my doubts.

 

[Bill Gates]

Any ETA on btrfs? I looked at their wiki and failed to notice any release dates?

It's already in the linux kernel.

 

[maflynn]

It works fine for people who use Linux - which tend to be more knowledgeable.
Would ext hold up to the rigors of the Poli Sci major fo example? I have my doubts.

Actually many people don't tinker with the file system. That's ok, if you have doubts, it neither impacts ext3/ext4's adoption or stability. I've had no problem with it running fedora and I can safely say I've done nothing with the file system that I haven't done with osx

 

[TechieJustin]

Actually many people don't tinker with the file system. That's ok, if you have doubts, it neither impacts ext3/ext4's adoption or stability. I've had no problem with it running fedora and I can safely say I've done nothing with the file system that I haven't done with osx

Your primary volume is HFS+ under Fedora?

 

[maflynn]

No, I mean ext3/ext4 (fedora 11 had ext3 and fedora 12 has ext4). Sorry, I was stating that I don't tinker (and I don't know anyone who tinkers) with the file system and its been solid and reliable w/o any problems. Early on ext4 had some issues that has been addressed so much so many of the current crop of linux distros default to ext4. I've not heard of any current problem with with filesytem.

 

[TechieJustin]

No, I mean ext3/ext4 (fedora 11 had ext3 and fedora 12 has ext4). Sorry, I was stating that I don't tinker (and I don't know anyone who tinkers) with the file system and its been solid and reliable w/o any problems. Early on ext4 had some issues that has been addressed so much so many of the current crop of linux distros default to ext4. I've not heard of any current problem with with filesytem.

Oh, ok.
What I meant was normal users tend to neglect their filesystem so it has to stand up to a certain amount of abuse - moreso than a mainframe. So my previous post wasn't meant to imply users wold intentionally tinker with the FS.