[besson3c]

HFS+

He cited the same argument we've heard in here about "it doesn't matter, as long as it does its job", and then shot it down by pointing out the several weaknesses and problems with it, and how it impacts regular users.

I just got to boot off of my Lion rescue disk for the first time since I needed to repair my disk using Disk Utility. The fact that Apple is so gung-ho about getting people to backup and keep their precious files on their Macs (Siracusa used iPhoto as an example) and HFS+ does nothing to ensure data integrity is a big problem. I'm tired of worrying about my directory structures exploding on me randomly. I recently had to run Disk Warrior on my wife's Powerbook (the hard drive on it has been recently replaced). The lack of data integrity just won't do.

He also provided many other interesting points and counter arguments to this idea, and glazed over the absence of other useful features in modern file systems: snapshots (which would be absolutely awesome with Lion, HFS+ is a huge bottleneck right now in requiring these crazy hard link directory structures), data deduplication, encryption (although it looks like Apple has done well with their improvements to Filevault using Core Storage), great software RAID features, compression, etc.

What was also interesting was that Siracusa also pontificated about whether Core Storage is laying the groundwork for a new file system. He, like me, seemed confused over what happened with the whole ZFS debacle. That still doesn't make any sense to me, and I'm not sure I really buy the while licensing thing as a comprehensive theory.

At any rate, HFS+ is a complete embarrassment. Lion is so awesome in so many ways, but the 25 year old HFS+ is just so out of place in OS X it's not even funny. I truly hope that Lion is the last version of OS X that I have to run HFS+.

[besson3c]

Sorry, rereading that I can see that it sounds like a rant, and I know that many of you bigger Apple fans are probably still a little giddy about Lion now, but let's just say that screwing around with rescue disks and DiskWarrior and wager is the opposite of fun.

[turtle777]

You are obviously reading John's article with your own prejudices.

Nowhere does John say it's the weakest part.

But where does this leave us on the file system front? Perhaps things are not as bad as they seem.

With ZFS out of the picture, Btrfs presumably eliminated due to its licensing, and future development of ReiserFS uncertain, its hard to see where Apple will get the modern file system that it so desperately needs other than by creating one itself.

This is it in a nutshell. There just isn't an alternative yet. Apple has to create one.
Sure, they could have been faster, but it's really not been the top priority so far.
And, to be honest, 99% of the users could care less.

-t

[besson3c]

Originally Posted by turtle777 

You are obviously reading John's article with your own prejudices.

Nowhere does John say it's the weakest part.

This is true, he doesn't say this explicitly, but this was my impression.

This is it in a nutshell. There just isn't an alternative yet. Apple has to create one.
Sure, they could have been faster, but it's really not been the top priority so far.
And, to be honest, 99% of the users could care less.

-t

99% of users couldn't care less because they don't understand the benefits of a better file system, but like I said, when they have data integrity, better performance, faster time machine backups, the ability to make local snapshots whenever they want, etc. they'll care.

[CharlesS]

Originally Posted by turtle777 

You are obviously reading John's article with your own prejudices.

Nowhere does John say it's the weakest part.

There's not really any question about it; technologically speaking, HFS+ is by far the most anachronistic part of OS X. The file system is every bit as creaky and duct-tapey as OS 9 was in its last days. The only positive things Siracusa has to say in the whole file system discussion is that there's hope that Apple might be writing a replacement, and that CoreStorage is a harbinger of that.

[besson3c]

What I'm also curious about is how Apple will ensure data integrity with their new file system, whatever that is.

In ZFS the file checksums exist on all disks in the array so that if a disk's checksum is off, it sends a copy from one of the other disks to replace the bad files. The checksums have to reside somewhere though, and you don't want them on the same drive on a single drive system since if this data goes bad you have lost all integrity.

Perhaps Apple will start shipping Macs with three built-in flash drives. If their new file system can handle RAID 5/6 or 10, they could get the capacity out of chaining very physically small Flash drives together, and if they copy what is done in ZFS the parity bits can live on all drives.

[lpkmckenna]

The many features added to HFS+ in recent years shows Apple is making the best of a bad situation. ZFS was to be their saviour, but the Sun/Oracle situation affected things, so Apple is certainly rolling their own solution now.

I think Apple is adopting their own tick-tock development process. Leopard was features, Snow Leopard was low-level, Lion is features again, and "Mountain Lion" will be low-level again, with a new file system.

[turtle777]

Why not Sea Lion ?

-t

[CharlesS]

Originally Posted by besson3c 

The checksums have to reside somewhere though, and you don't want them on the same drive on a single drive system since if this data goes bad you have lost all integrity.

Yes, but at least you'd know about it, and you'd know that it's time to restore from backup. Or, if it's your backup drive that has the checksum mismatch, you'll know that it's time to make a new backup, and that it's definitely not a good time to, say, reformat your main drive and restore from the backup.

Knowing that your data has or hasn't been compromised is not a bad thing to have.

[angelmb]

Originally Posted by turtle777 

Why not Sea Lion ?

-t

Aqua no more ?

[OreoCookie]

@besson
First of all, you don't need to convince me how sh*tty HFS+ is, I agree with you on all counts. I like ZFS very much and I was very disappointed when Apple canned its migration to ZFS for what seems to be intellectual property reasons. OS X is in desperate need of a modern file system.

Developing a new file system is hard work and I have a hard time believing that any modern file system built from the ground up won't violate similar patents than ZFS or BTRFS. Given how much cash Apple has in its bank, I was a bit surprised: surely, the combined might of Oracle and Apple* should be enough to scare or pay off anyone who tries to challenge them for IP infringement.

But from an end-user perspective, does it matter? I don't know how, but Apple managed to add versions to the file system, it managed to do Time Machine, the end user doesn't care whether the files are stored on paper cards, HFS+ or ZFS.

* It should be noted that Ellison and Jobs are good friends, so I don't think it should be difficult to hammer out a license agreement for ZFS if Jobs wants to.

[Big Mac]

Based on the reviews and user critiques I've read, HFS+ is the least of Lion's weakest parts. Although I'm impressed by the apparent lack of any serious bugs or incompatibilities (like many other 10.X.0 releases had), it sounds like the interface changes were rushed and poorly planned.

HFS+ may be lacking modern FS features, but for all its perceived and discussed flaws it has done little to impede OS X adoption over the last decade. I think it's a considerable exaggeration to claim it's such a major albatross.

[besson3c]

Originally Posted by Big Mac 

Based on the reviews and user critiques I've read, HFS+ is the least of Lion's weakest parts. Although I'm impressed by the apparent lack of any serious bugs or incompatibilities (like many other 10.X.0 releases had), it sounds like the interface changes were rushed and poorly planned.

HFS+ may be lacking modern FS features, but for all its perceived and discussed flaws it has done little to impede OS X adoption over the last decade. I think it's a considerable exaggeration to claim it's such a major albatross.

You are conflating OS X's market success with its technical merits. There are probably more Windows XP machines out there than Macs, does that make Windows XP better?

[angelmb]

it sounds like the interface changes were rushed and poorly planned.

I guess that refers to iCal and Address Book, cause the whole system looks so classy. Reminds me of Mac OS 8 heydays.

[P]

The thing about ZFS is that it's not really a perfect fit for Apple. If it's available quickly and cheaply, then great - it's so much better than HFS+ that's not even funny - but if it's hard or expensive, Apple could make a better FS internally. My guess is that Oracle simply wanted to keep ZFS to itself after the purchase, and Apple wasn't willing to pay through the nose for a filesystem that is not a good fit.

At this point, I have to wonder about patents - at least if you don't aim for ZFS-quality. XFS came out in 1994, so any patents for that would expire in the 3 years max. A filesystem of XFS quality would be an enormous step up from HFS+ - why not just try for that?

[OreoCookie]

Originally Posted by P 

The thing about ZFS is that it's not really a perfect fit for Apple.

In what way? Granted, ZFS includes a lot of stuff the average Mac user doesn't need, but I could, for instance, think of hybrid storage solutions (RAM + SSD + platter-based hard drive) to speed up many disk operations.

I guess the biggest reason why people tend focus on ZFS is because there isn't much else out there in terms of a modern filesystem. There's BTRFS, but butter FS is less mature and suffers from the lethal stigma called GPL. (The GPL isn't bad per se, it's just incompatible with Apple's open source license.)

Originally Posted by P 

At this point, I have to wonder about patents - at least if you don't aim for ZFS-quality. XFS came out in 1994, so any patents for that would expire in the 3 years max. A filesystem of XFS quality would be an enormous step up from HFS+ - why not just try for that?

I'm not sure whether replacing 30-year old technology with 20-year old technology is a good idea when you want something that is cutting edge. Apple has the resources to develop its own file system, though, and I hope this is the road they're taking. They've hired good people (e. g. the inventor of BeFS), but so far, nothing has surfaced (contrary to my own expectations).

[cgc]

Originally Posted by P 

...At this point, I have to wonder about patents - at least if you don't aim for ZFS-quality. XFS came out in 1994, so any patents for that would expire in the 3 years max. A filesystem of XFS quality would be an enormous step up from HFS+ - why not just try for that?

I'm sure Lodsys is camping on some random patents they can relate to ZFS and is drooling waiting for Apple to use it </sarcasm>

[msuper69]

In 13 years of using Macs, I can count the number of problems I've had with the file system on 1 finger and that was caused by a power outage (before I had a UPS and before journaling was available.

I'm not a typical Mac user but I just don't see this as a big problem in the grand scheme of things.

[OreoCookie]

@msuper
It's not just about data safety, but also about performance and the types of features you can add. Filesystem-level snapshots, for instance, could be used to be able to roll back software updates, increase the speed of Time Machine backups and make the use of versions more efficient.

ZFS can use SSDs (also cheap SSDs) and RAM to speed up reading/writing. Think of Apple including a relatively small SSD in each Mac and you get SSD-type speeds of most files. (ZFS is quite smart, it can cache the most frequently accessed and most recent files in RAM and use the SSD to cover up the `write hole'.)

[cgc]

Originally Posted by msuper69 

In 13 years of using Macs, I can count the number of problems I've had with the file system on 1 finger and that was caused by a power outage (before I had a UPS and before journaling was available.

I'm not a typical Mac user but I just don't see this as a big problem in the grand scheme of things.

I feel the same way but could it be we are so naive regarding the benefits of a "modern" FS we don't know what we don't know? HFS+ is good, never had a problem I could trace back to HFS+ failing but could ZFS, BeFS, or XFS bring some improvement we couldn't foresee?

Would ext4 be much of an improvement?

[cgc]

Originally Posted by OreoCookie 

@msuper
It's not just about data safety, but also about performance and the types of features you can add. Filesystem-level snapshots, for instance, could be used to be able to roll back software updates, increase the speed of Time Machine backups and make the use of versions more efficient.

ZFS can use SSDs (also cheap SSDs) and RAM to speed up reading/writing. Think of Apple including a relatively small SSD in each Mac and you get SSD-type speeds of most files. (ZFS is quite smart, it can cache the most frequently accessed and most recent files in RAM and use the SSD to cover up the `write hole'.)

Uh huh huh huh...you said "write hole"

[Spheric Harlot]

Originally Posted by OreoCookie 

Apple has the resources to develop its own file system, though, and I hope this is the road they're taking. They've hired good people (e. g. the inventor of BeFS), but so far, nothing has surfaced (contrary to my own expectations).

A file system is one of those things that really, really, REALLY need to work absolutely 100% perfectly before you build your system around them.

[OreoCookie]

Originally Posted by cgc 

Would ext4 be much of an improvement?

Not really: ext4 is a souped-up version of ext3 which itself is essentially ext2 + journaling -- and ext2 is 20 years old. It is intended to be a stop-gap solution up until btrfs becomes stable enough to be considered as default file system for Linux.*

* The history here is quite interesting: there already was a very successful file system that people thought would supplant ext2 and ext3, ReiserFS. The problem with ReiserFS was two-fold: apparently, the main developer was difficult to work with and worse, the main developer was convicted for murder. So it's not quite clear whether ReiserFS 4 will ever make it into the kernel.

[besson3c]

Originally Posted by lpkmckenna 

The many features added to HFS+ in recent years shows Apple is making the best of a bad situation. ZFS was to be their saviour, but the Sun/Oracle situation affected things, so Apple is certainly rolling their own solution now.

It must have, but there was a ton of ZFS code that was already open sourced. Apple could have just become the unofficial overlord of ZFS similar to how they became the overlord of KHTML.

[besson3c]

Originally Posted by CharlesS 

Yes, but at least you'd know about it, and you'd know that it's time to restore from backup. Or, if it's your backup drive that has the checksum mismatch, you'll know that it's time to make a new backup, and that it's definitely not a good time to, say, reformat your main drive and restore from the backup.

Knowing that your data has or hasn't been compromised is not a bad thing to have.

That's true, this would be much better than the silent failures we have today.

[besson3c]

Originally Posted by OreoCookie 

But from an end-user perspective, does it matter? I don't know how, but Apple managed to add versions to the file system, it managed to do Time Machine, the end user doesn't care whether the files are stored on paper cards, HFS+ or ZFS.

As has been said a number of times thus far, they will care when they can't boot their machine and they need to find somebody that has a copy of Disk Warrior and a rescue disk and all that jazz.

They will also care when they see what kind of performance gains can be had with a modern file system, particularly with Time Machine.

[besson3c]

Originally Posted by P 

The thing about ZFS is that it's not really a perfect fit for Apple. If it's available quickly and cheaply, then great - it's so much better than HFS+ that's not even funny - but if it's hard or expensive, Apple could make a better FS internally. My guess is that Oracle simply wanted to keep ZFS to itself after the purchase, and Apple wasn't willing to pay through the nose for a filesystem that is not a good fit.

At this point, I have to wonder about patents - at least if you don't aim for ZFS-quality. XFS came out in 1994, so any patents for that would expire in the 3 years max. A filesystem of XFS quality would be an enormous step up from HFS+ - why not just try for that?

Is free cheap enough?

The bulk of ZFS was open sourced. The only parts that were proprietary were stuff relating to encryption and stuff like that.

[turtle777]

Originally Posted by besson3c 

Is free cheap enough?

The bulk of ZFS was open sourced. The only parts that were proprietary were stuff relating to encryption and stuff like that.

Why then does John make it sound like licensing was such a big deal ?

-t

[besson3c]

Originally Posted by OreoCookie 

In what way? Granted, ZFS includes a lot of stuff the average Mac user doesn't need, but I could, for instance, think of hybrid storage solutions (RAM + SSD + platter-based hard drive) to speed up many disk operations.

This is a great point that I forgot to include in my list.

The RAM/SSD buffer thing will increase performance of stuff in OS X *big time*. If, say, the guts of your favorite application have been cached into RAM or a fast SSD, we are talking about instant app launches and stuff.

So, please, please, please, please stop saying that ordinary users should not care folks. Nothing is further from the truth. Ordinary users don't understand this stuff and don't think they'll benefit because they don't understand this stuff, but *EVERYBODY* will benefit from a new modern file system and the capabilities that go along with it.

[besson3c]

Originally Posted by OreoCookie 

@msuper
It's not just about data safety, but also about performance and the types of features you can add. Filesystem-level snapshots, for instance, could be used to be able to roll back software updates, increase the speed of Time Machine backups and make the use of versions more efficient.

ZFS can use SSDs (also cheap SSDs) and RAM to speed up reading/writing. Think of Apple including a relatively small SSD in each Mac and you get SSD-type speeds of most files. (ZFS is quite smart, it can cache the most frequently accessed and most recent files in RAM and use the SSD to cover up the `write hole'.)

I repeat: we will *all* benefit from a modern file system. Period.

[lpkmckenna]

@turtle777: ZFS was doomed to abandonment under Oracle because of their own BTRFS. Buying into a doomed platform looked like a bad idea to Apple.

[besson3c]

Originally Posted by cgc 

Would ext4 be much of an improvement?

Yes, but ext4 is sort of a bridge between the old to the new. It is better than HFS+ (as is pretty much everything out there including even possibly NTFS), but it probably makes sense for Apple to put their time into getting the platform into a file system that will give them room to grow and that will be around for years.

With ZFS for instance, the file system is extensible and versioned, so you can tack on new things and apply these new things to new systems via a download/install.

I hope that Apple opens up their file system if they are making one internally. It will probably accelerate its improvement and stability.

[OreoCookie]

Originally Posted by turtle777 

Why then does John make it sound like licensing was such a big deal ?

According to the grapevine the reason Apple decided not to choose ZFS as its next-gen file system was the unclear IP situation, i. e. whether adopting ZFS would make Apple vulnerable for being sued for patent infringement.
On the one hand, these are just rumors and we have no way of checking whether this is actually true, but on the other hand it's a rumor that has been spread consistently and frankly, there are really no other good reasons people could come up with. (ZFS support was removed very late and after being officially announced as a feature of OS X server. There was an official ZFS SDK for developers to test.)

Now that Oracle has vowed to continue support of both, ZFS and btrfs, I have no idea what's keeping Apple from trying ZFS again. Perhaps they have their own skunkworks project now?

[besson3c]

Originally Posted by turtle777 

Why then does John make it sound like licensing was such a big deal ?

-t

Because this seems to be the popular narrative? I don't know...

Here's what I know:

The bulk of ZFS is open source. There are some proprietary bits, but not enough to prevent adaptation as the FreeBSD team has done with a fraction of Apple's resources. Apparently there have been some Apple reps at the Illumos org. Illumos - Wikipedia, the free encyclopedia

Maybe Apple was concerned over how Illumos was playing out, or what the relationship between the parent company Oracle would be like, who controls what and how, politics, etc.

Still, I don't understand why Apple could have avoided that by creating their own ZFS fork similar to how they did so with Webkit and KHTML. They didn't need to get involved with any of this messy stuff, they could have just taken the code that was available and put it to use. AFAIK the CDDL is totally compatible with Apple: Common Development and Distribution License - Wikipedia, the free encyclopedia

[besson3c]

Originally Posted by lpkmckenna 

@turtle777: ZFS was doomed to abandonment under Oracle because of their own BTRFS. Buying into a doomed platform looked like a bad idea to Apple.

But did this matter? Like I said, Apple could have just created their own fork.

[turtle777]

*shrug*

I doubt it was w/o a good reason.

-t

[besson3c]

Originally Posted by turtle777 

*shrug*

I doubt it was w/o a good reason.

-t

I doubt it was a good reason. As Siracusa pointed out, this whole debacle has really set Apple back, cost them significant amount of time. If these issues did exist, could they have not been identified prior to Apple starting the ZFS OS X port?

[turtle777]

Well, carry on thinking Apple is a lazy, mean-spirited bunch that could have easily whipped up a new file system, but just chose to not do so for now reason whatsoever.

-t

[besson3c]

Originally Posted by turtle777 

Well, carry on thinking Apple is a lazy, mean-spirited bunch that could have easily whipped up a new file system, but just chose to not do so for now reason whatsoever.

-t

Are you okay, Turtle? Where on Earth did this come from? I didn't say this, I said that Apple probably just screwed this up. They are subject to human error, you know.

[turtle777]

So you believe that Apple just screwed it up ?
Ok, do that and carry on.

I think it's far more likely that there are GOOD reasons.

-t

[besson3c]

Originally Posted by turtle777 

So you believe that Apple just screwed it up ?
Ok, do that and carry on.

I think it's far more likely that there are GOOD reasons.

-t

There could be, but whatever the reason is they still screwed all of this up by wasting time going down a path that ended up being a dead end, and they screwed this up by not having a new file system ready for at least beta testing in Lion for advanced users that are okay with formatting a drive with this new file system for putting it through its paces.

[turtle777]

What if the actually did the RIGHT thing by delaying this ?
Imagine: they could have rushed out a broken new filesystem and made things WORSE.

Since we don't have any details, speculation that Apple just screwed it up is useless.

-t

[besson3c]

Originally Posted by turtle777 

What if the actually did the RIGHT thing by delaying this ?
Imagine: they could have rushed out a broken new filesystem and made things WORSE.

Since we don't have any details, speculation that Apple just screwed it up is useless.

-t

Embarking down a path on a project and then puling the plug on this means a mistake was made. I'm not saying it was a bad or good decision, but that same time that was spent allocating resources to the ZFS for OS X project could have been spent on whatever else had they known that ZFS for OS X was going to be a dead end.

Hence, Apple made at least one mistake.

[msuper69]

Originally Posted by OreoCookie 

@msuper
It's not just about data safety, but also about performance and the types of features you can add. Filesystem-level snapshots, for instance, could be used to be able to roll back software updates, increase the speed of Time Machine backups and make the use of versions more efficient.

ZFS can use SSDs (also cheap SSDs) and RAM to speed up reading/writing. Think of Apple including a relatively small SSD in each Mac and you get SSD-type speeds of most files. (ZFS is quite smart, it can cache the most frequently accessed and most recent files in RAM and use the SSD to cover up the `write hole'.)

Yeah, I wasn't thinking of the file system's other capabilities. DUH.

[P]

Originally Posted by OreoCookie 

In what way? Granted, ZFS includes a lot of stuff the average Mac user doesn't need, but I could, for instance, think of hybrid storage solutions (RAM + SSD + platter-based hard drive) to speed up many disk operations.

That doesn't strictly have to be in the filesystem, though. Intel has a solution for bolting SSD caching (OS X already has very good RAM caching) onto any filesystem under Windows using bastardized RAID1, and MS also added the same thing on top of NTFS in Vista.

Originally Posted by OreoCookie 

I guess the biggest reason why people tend focus on ZFS is because there isn't much else out there in terms of a modern filesystem. There's BTRFS, but butter FS is less mature and suffers from the lethal stigma called GPL. (The GPL isn't bad per se, it's just incompatible with Apple's open source license.)

I agree with all this, but my point is that what Apple needs is not necessarily met by ZFS. They need better performance, they need to get rid of the worst hacks used to implement Spotlight and Time Machine (hard links, snapshots) and they need better threading. There are many filesystems that can do that that are ZFS and its potentially patent-infringing goodness.

Originally Posted by OreoCookie 

I'm not sure whether replacing 30-year old technology with 20-year old technology is a good idea when you want something that is cutting edge. Apple has the resources to develop its own file system, though, and I hope this is the road they're taking. They've hired good people (e. g. the inventor of BeFS), but so far, nothing has surfaced (contrary to my own expectations).

The filesystem group at Apple is quite small, and it managed to produce HFS+ journalling, partition resizing, all those hacks for Time Machine, and Spotlight. Spotlight really is pretty awesome - it even lets you do network searches, which I don't think anyone else does. I did hope for something great for Lion, but so far the only thing I see is the volume manager stuff.

[besson3c]

Originally Posted by P 

That doesn't strictly have to be in the filesystem, though. Intel has a solution for bolting SSD caching (OS X already has very good RAM caching) onto any filesystem under Windows using bastardized RAID1, and MS also added the same thing on top of NTFS in Vista.

The ZFS stuff is not just caching though, it's also buffer so that the disks aren't stressed as hard. This allows you to use cheap SATA disks for permanent storage, and fast devices for handling requests.

I agree with all this, but my point is that what Apple needs is not necessarily met by ZFS. They need better performance, they need to get rid of the worst hacks used to implement Spotlight and Time Machine (hard links, snapshots) and they need better threading. There are many filesystems that can do that that are ZFS and its potentially patent-infringing goodness.

There is no file system out there that offers the totality of ZFS though, as well as all of the features we have touched on: dedup, encryption, all sorts of different RAID modes, read/write buffers assigned to disks of your choice, snapshots, automatic compression, NFS and iSCSI sharing, copy on write, manual scrubbing, live RAID config changes, etc. some of these features would be more useful than others in OS X, but if Apple hasn't even gotten a beta file system out the door it looks like they have a lot of work to do if they want to catch up to ZFS.

Benefits:

- dedup: less disk space consumed
- encryption: move existing encryption to a lower level
- RAID modes: decisions on level of redundancy in server environments (I guess this would probably be irrelevant to Apple, for now)
- buffers: as mentioned above
- snapshots: time machine backups in literally a second or two, no exaggeration, no hard link mess
- compression: less disk space consumed, faster performance in trading off I/O for CPU
- NFS/iSCSI sharing: perhaps not all that useful to Apple since they are into AFP
- copy on write: data integrity, no more DiskWarrior/Disk Utility repairs/fsck
- manual scrubbing: same

[OreoCookie]

Originally Posted by P 

That doesn't strictly have to be in the filesystem, though. Intel has a solution for bolting SSD caching (OS X already has very good RAM caching) onto any filesystem under Windows using bastardized RAID1, and MS also added the same thing on top of NTFS in Vista.

I realize that most of the functionality can be `just bolted on', SSD caching would just be another in a long list Apple could put on top of HFS+ (along with Time Machine backups, Spotlight, etc.). But the thing is that ZFS was designed with this stuff in mind -- as would in all likelihood be any alternative to ZFS.

Originally Posted by P 

I agree with all this, but my point is that what Apple needs is not necessarily met by ZFS. They need better performance, they need to get rid of the worst hacks used to implement Spotlight and Time Machine (hard links, snapshots) and they need better threading. There are many filesystems that can do that that are ZFS and its potentially patent-infringing goodness.

I agree.
But I don't see a viable candidate on the radar. Given Apple's track record with technologies, I don't think they'll go for some 20-year old file system such as XFS, just because it is `safe'. What do you see as the most likely candidates to succeed HFS+?

[besson3c]

Originally Posted by OreoCookie 

I realize that most of the functionality can be `just bolted on', SSD caching would just be another in a long list Apple could put on top of HFS+ (along with Time Machine backups, Spotlight, etc.). But the thing is that ZFS was designed with this stuff in mind -- as would in all likelihood be any alternative to ZFS.

I agree.
But I don't see a viable candidate on the radar. Given Apple's track record with technologies, I don't think they'll go for some 20-year old file system such as XFS, just because it is `safe'. What do you see as the most likely candidates to succeed HFS+?

I think if Apple weren't going the homegrown route they'd probably be able to get something out the door now, but given that there are no signs of anything it is probably safe to say that it is going to be a very long time before a new file system is deemed as production worthy. My wishing for the next major release is probably wishful thinking.

[P]

Originally Posted by OreoCookie 

I agree.
But I don't see a viable candidate on the radar. Given Apple's track record with technologies, I don't think they'll go for some 20-year old file system such as XFS, just because it is `safe'. What do you see as the most likely candidates to succeed HFS+?

1) HFS3, ie a homegrown replacement. Given the talent currently at Apple, that could be just awesome.
2) XFS. I have some experience with it from back when I worked with SGI machines in university, and it's quite powerful and solves the performance issue. I also think that it can quite easily be improved by Apple to be updated to modern consideration. I really think that SSD caching should be handled in another layer (ie, be block-based rather than file-based), so that is not an issue. If I have managed to trace the assets correctly, it s currently owned by the SGI successor company (formerly Rackspace) which has a market cap of 500 million. Even with a decent premium, Apple could buy it and all patents for a song (comparatively).
3) ZFS. This would mean giving up on some other plan, and Apple rarely does that, but it's a possibility - they could buy Ten's Complement.
4) BtrFS or JFS2. Either would require a deal with another big company (Oracle or IBM), and neither of them have anything to gain on such a deal, but it's an interesting idea.

[P]

Originally Posted by besson3c 

I think if Apple weren't going the homegrown route they'd probably be able to get something out the door now, but given that there are no signs of anything it is probably safe to say that it is going to be a very long time before a new file system is deemed as production worthy. My wishing for the next major release is probably wishful thinking.

What, because Apple is so bad at keeping secrets? They could put it in 10.7.2 if they'd like - remember that journalling was added in a point release, and HFS+ itself was added in a free release (8.1).