[TETENAL]

http://en.wikipedia.org/wiki/ZFS

Apple seems to be interested in including ZFS into OS X. According to Comparison of file systems at Wikipedia it only supports 255 bytes filenames while HFS+ supports 255 UTF-16 characters. ZFS is also case sensitive. So we can only hope this will not become the default file system in the future.

http://www.opensolaris.org/os/commun.../demos/basics/

[Person Man]

Originally Posted by TETENAL

http://en.wikipedia.org/wiki/ZFS

Apple seems to be interested in including ZFS into OS X. According to Comparison of file systems at Wikipedia it only supports 255 bytes filenames while HFS+ supports 255 UTF-16 characters. ZFS is also case sensitive. So we can only hope this will not become the default file system in the future.

http://www.opensolaris.org/os/commun.../demos/basics/

You would reject it as a default file system simply on the basis that Unicode file names would be limited to 127 or 128 characters?

Perhaps Apple will extend it such that it provides all the features of the file system plus the ability to store 255 Unicode characters.

[TETENAL]

Originally Posted by Person Man

You would reject it as a default file system simply on the basis that Unicode file names would be limited to 127 or 128 characters?

What happens when you copy a file with a name longer than 127 characters from a HFS+ to a ZFS volume? What happens when you copy a folder containing files with the same name different only in capitalization from a ZFS volume to a HFS+ volume? ZFS would cause a lot of trouble for the user. The advantages on the other hand are of obscure nature.

What I would like to see in the filesystem is a per file encryption, so that we can do without the FileVault hack. Wikipedia doesn't say whether ZFS supports that or not.

[Person Man]

Originally Posted by TETENAL

What happens when you copy a file with a name longer than 127 characters from a HFS+ to a ZFS volume? What happens when you copy a folder containing files with the same name different only in capitalization from a ZFS volume to a HFS+ volume? ZFS would cause a lot of trouble for the user. The advantages on the other hand are of obscure nature.

Well, what happens when you copy a case-sensitive file from UFS to case-insensitive HFS+? And there is a case-sensitive HFS plus format option now, too.

Who is to say that Apple's implementation of ZFS won't support 255 Unicode character names or case insensitivity (as an option)?

It's way too early to pass judgement just yet.

[Catfish_Man]

ZFS is designed to be extensible; 255 character filename support could be added (although presumably at a slight cost in performance).

[Tuoder]

ZFS is fast, and supports drives of effectively any size. That is why they like it.

[OreoCookie]

ZFS would be a tremendous leap forward. I took a bit of a closer look last night and it seems it's the most advanced filesystem I've seen so far.

I don't think the 255 byte vs. 255 Unicode characters issue will be such a big issue for most of us. It doesn't make a difference unless you use Asian languages (Japanese and Chinese in particular).

I'm also not sure what you mean by `ZFS is case-sensitive': so what? HFS+ is case-sensitive as well (if you install OS X server, you can choose betwenn HFS+ case-sensitive and HFS+ case-preserving. In any case, case-sensitivity is a big plus and not an issue anymore on a Mac.

Also if Apple puts a sort of spotlight-like architecture on top of zfs, you could have filenames of `arbitrary' length as metadata.

I really do hope ZFS will be implemented in one form or another in the future.

[tooki]

Whoa there, pardner. Case-sensitivity is not categorically "a big plus". It's a nightmare for users, who then have to understand that "My File" and "My file" are entirely different. (Think about having to explain this to your mom.) Apple chose to make a case-insensitive, case-preserving filesystem for a reason -- it takes a lot more effort to implement, you know --, and only added the case-sensitive version as an option for those who need that support for applications ported from other operating systems (it's not just in OS X Server, by the way). I would hate to see us move backwards to a case-sensitive filesystem.

That said, I agree that it might be possible and efficient to just modify a high-tech filesystem like ZFS to support better file names and case-insensitivity rather than rolling your own.

tooki

[P]

The only thing we know is that someone at Apple (the filesystem support developer) is looking to port it. Having a port of a file system does NOT mean that OS X will support rooting and booting from it - compare the support for FAT32 in today's OS.

The implementation of the kernel interfaces in Tiger means that it is finally feasible to start developing such filesystem plug-ins without having to update them for every minor revision. If I were a file system developer and wanted to test this system as well as promote it a bit, I'd start by developing a plug-in for the most advanced file system I could find. Any bugs I found or any functions missing from the kernel interface discovered during this implementation could then be added in 10.4.9 or 10.5 or whatever and make the model that much more complete. One of the promises with xnu was that filesystem plug-ins like this would be easy to write. So far, there is only one third party plugin that is still supported (an efs2 plug) and another (NTFS) that was abandoned when Apple added that support to the OS. I'm sure Apple would like to see plug-ins for every filesystem out there added as second-tier filesystems, and with XFS, JFS and the Reisers being under the GPL porting them would be a snap if anyone knew how and could trust the kernel interfaces.

[Person Man]

Originally Posted by P

The only thing we know is that someone at Apple (the filesystem support developer) is looking to port it. Having a port of a file system does NOT mean that OS X will support rooting and booting from it - compare the support for FAT32 in today's OS.

The implementation of the kernel interfaces in Tiger means that it is finally feasible to start developing such filesystem plug-ins without having to update them for every minor revision. If I were a file system developer and wanted to test this system as well as promote it a bit, I'd start by developing a plug-in for the most advanced file system I could find. Any bugs I found or any functions missing from the kernel interface discovered during this implementation could then be added in 10.4.9 or 10.5 or whatever and make the model that much more complete. One of the promises with xnu was that filesystem plug-ins like this would be easy to write. So far, there is only one third party plugin that is still supported (an efs2 plug) and another (NTFS) that was abandoned when Apple added that support to the OS. I'm sure Apple would like to see plug-ins for every filesystem out there added as second-tier filesystems, and with XFS, JFS and the Reisers being under the GPL porting them would be a snap if anyone knew how and could trust the kernel interfaces.

Yes, but the whole point of implementing ZFS is to support "rooting and booting" from it because of its efficiency.

[Person Man]

Originally Posted by tooki

That said, I agree that it might be possible and efficient to just modify a high-tech filesystem like ZFS to support better file names and case-insensitivity rather than rolling your own.

tooki

Yes. ZFS is meant to be extensible, so I'd bet extending the number of allowable characters and making a case-insensitive option could be done.

Edit: 2000 posts. w00t!

[OreoCookie]

Originally Posted by tooki

Whoa there, pardner. Case-sensitivity is not categorically "a big plus". It's a nightmare for users, who then have to understand that "My File" and "My file" are entirely different. (Think about having to explain this to your mom.) Apple chose to make a case-insensitive, case-preserving filesystem for a reason -- it takes a lot more effort to implement, you know --, and only added the case-sensitive version as an option for those who need that support for applications ported from other operating systems (it's not just in OS X Server, by the way). I would hate to see us move backwards to a case-sensitive filesystem.

It's a move forward, not a move backwards. Some unix apps won't compile properly because of that problem for instance. And the thing is that HFS+ can do both, so it doesn't seem to be that much of an effort programming-wise.

In any case, you could still choose to have the best of both worlds with meta data (e. g. you could have a field `long file name' which is case-preserving).

Originally Posted by tooki

That said, I agree that it might be possible and efficient to just modify a high-tech filesystem like ZFS to support better file names and case-insensitivity rather than rolling your own.

The file name length thing is largely a theoretical advantage, so even if nothing changes here, you wouldn't notice the difference unless you can speak Chinese or Japanese.

[TETENAL]

Originally Posted by OreoCookie

It's a move forward, not a move backwards. Some unix apps won't compile properly because of that problem for instance.

Case sensitivity is a huge step backwards in usability. It might make it easier to compile command line applications from the 60s, but the Mac should put the user first, not the programmer. So case insensitivity is better (The Mac kind of better not the other kind you've been thinking of).

Originally Posted by OreoCookie

In any case, you could still choose to have the best of both worlds with meta data (e. g. you could have a field `long file name' which is case-preserving).

Since all file systems are case preserving that is not an issue.

Originally Posted by OreoCookie

The file name length thing is largely a theoretical advantage, so even if nothing changes here, you wouldn't notice the difference unless you can speak Chinese or Japanese.

Which makes quite a lot people.

[tooki]

Originally Posted by OreoCookie

It's a move forward, not a move backwards. Some unix apps won't compile properly because of that problem for instance. And the thing is that HFS+ can do both, so it doesn't seem to be that much of an effort programming-wise.

No, case-sensitivity is a move backwards. Case-sensitivity is the easy way out for the programmer: just compare bytes. Case-insensitivity is the human (and humane!) option.

UNIX apps that won't compile should be fixed. That's a better solution than forcing users to remember case. (And for right now, the option to use a case-sensitive filesystem is there for apps that haven't yet been fixed.)

And yeah, Unicode file names are very important. You realize that Asians use a lot of computers, right?

tooki

[OreoCookie]

Originally Posted by tooki

And yeah, Unicode file names are very important. You realize that Asians use a lot of computers, right?

Take a look at my (out-of-date) location
You don't have to tell me

Personally, I very much prefer precision. If I can have it, I want to have it (after all, half of me is a mathematician).

[tooki]

For most people, case-sensitivity goes beyond precision into nitpicking. An "a" is an "A" is an "a", after all. For usability, case-sensitivity is a nightmare. Besides, it's not as if HFS+ destroys cases: it does preserve them, it merely treats case the same way humans do, rather than the way computers do.

tooki

[Salty]

If it makes my PowerBook faster...

[Madrag]

I think apple will take all the points above into account and develop a "better" ZFS.
I really like the self-healing feature of ZFS!

[Tuoder]

Originally Posted by Madrag

I think apple will take all the points above into account and develop a "better" ZFS.
I really like the self-healing feature of ZFS!

I forgot about that. That sounded very interesting when I originally read about it. I wonder how exactly that works.

[OreoCookie]

Originally Posted by tooki

For most people, case-sensitivity goes beyond precision into nitpicking. An "a" is an "A" is an "a", after all. For usability, case-sensitivity is a nightmare. Besides, it's not as if HFS+ destroys cases: it does preserve them, it merely treats case the same way humans do, rather than the way computers do.

Humans do treat cases more seriously. Take German for instance: all nouns are in capital letters. `Gut' and `gut' even differ in meaning and not just a difference between one being a noun and the other being an adjective. So saying humans do not distinguish between cases is incorrect.

More importantly, when does it make a difference to the end user whether his filesystem is case-preserving or case-sensitive? Only if the user tries to create several files with names that differ only by cases. And if you `do not distinguish' between cases, you wouldn't do that anyway. Otherwise you would end up with two files that have `the same' name.

On the other hand, some people/applications do care.

[Horsepoo!!!]

Originally Posted by OreoCookie

Humans do treat cases more seriously. Take German for instance: all nouns are in capital letters. `Gut' and `gut' even differ in meaning and not just a difference between one being a noun and the other being an adjective. So saying humans do not distinguish between cases is incorrect.

More importantly, when does it make a difference to the end user whether his filesystem is case-preserving or case-sensitive? Only if the user tries to create several files with names that differ only by cases. And if you `do not distinguish' between cases, you wouldn't do that anyway. Otherwise you would end up with two files that have `the same' name.

On the other hand, some people/applications do care.

It'd be difficult to justify the different behaviors. If there isn't one and only one behavior, people will get confused when they switch from one computer that allows case-sensitive behaviors and one that doesn't. It would also cause a problem on networks. About 99% of people are used to a case-insensitive file system.

So it's really hard to tell how seriously humans treat cases. Nobody said they can't distinguish between cases...that would be a pretty silly thing to say. But there are some that don't pay a whole lot of attention to cases...especially in a particular context such as naming files on a computer.

Case-insensitiveness exists to remove any confusion...and frankly, it does...end of case. Adding a case-preserving option would add unnecessary complexity to satisfy 1% of people out there? (and I'm being generous with that 1%)

[OreoCookie]

Originally Posted by Horsepoo!!!

It'd be difficult to justify the different behaviors. If there isn't one and only one behavior, people will get confused when they switch from one computer that allows case-sensitive behaviors and one that doesn't. It would also cause a problem on networks. About 99% of people are used to a case-insensitive file system.

All new filesystems are case-sensitive and that includes NTFS. Only FATxx filesystems are case-insensitive and still used. So losing case-sensitivity would add problems, not the other way around. And in the end, only the case-insensitive computer would end up with the problems.

Originally Posted by Horsepoo!!!

So it's really hard to tell how seriously humans treat cases. Nobody said they can't distinguish between cases...that would be a pretty silly thing to say. But there are some that don't pay a whole lot of attention to cases...especially in a particular context such as naming files on a computer.

This is something that comes from your native language, English, but which is not true for others.

Originally Posted by Horsepoo!!!

Case-insensitiveness exists to remove any confusion...and frankly, it does...end of case. Adding a case-preserving option would add unnecessary complexity to satisfy 1% of people out there? (and I'm being generous with that 1%)

I don't think you have any measure of telling how many people would need case-sensitivity. Also, you still haven't explained why having a case-sensitive filesystem adds confusion when you typically don't have two files with almost identical names?

By the way, HFS+ is case-preserving which some people want to keep (and whose side I presume you are arguing for).

[Horsepoo!!!]

Originally Posted by OreoCookie

All new filesystems are case-sensitive and that includes NTFS. Only FATxx filesystems are case-insensitive and still used. So losing case-sensitivity would add problems, not the other way around. And in the end, only the case-insensitive computer would end up with the problems.

This is something that comes from your native language, English, but which is not true for others.

Except that my native language is...French. And wtg on showing the exact contrary in that same sentence.

I don't think you have any measure of telling how many people would need case-sensitivity. Also, you still haven't explained why having a case-sensitive filesystem adds confusion when you typically don't have two files with almost identical names?

But if they did, there could be confusion.

By the way, HFS+ is case-preserving which some people want to keep (and whose side I presume you are arguing for).

Yeah...sorry, preserving, yes...sensitive, no, unless you're ready to explain to a user why 'document.txt.' and 'Document.txt' aren't the same file. If you're ready to take on this task then I don't mind that file systems become case-sensitive.

[ksloan2]

Originally Posted by Salty

If it makes my PowerBook faster...

Well said. Apple will sort the case-sensitivity issues out, I'm not interesting in debating that. All that matters is speed and added functionality.

[OreoCookie]

Originally Posted by Horsepoo!!!

Except that my native language is...French. And wtg on showing the exact contrary in that same sentence.

That doesn't invalidate what I've said. All modern filesystems are case-sensitive and all that are in use (with the exception of HFS+ for OS X client users). And in French most of the words are lower-case, so you're not trained to look for the difference.

Originally Posted by Horsepoo!!!

But if they did, there could be confusion.

If they do, they do it on purpose. So no, there is no confusion. Usually you don't create two files with names that differ by case, but if you do, you know why. And that it makes a difference.

So all the people who prefer case-preserving filesystems wouldn't feel a difference.

Originally Posted by Horsepoo!!!

Yeah...sorry, preserving, yes...sensitive, no, unless you're ready to explain to a user why 'document.txt.' and 'Document.txt' aren't the same file. If you're ready to take on this task then I don't mind that file systems become case-sensitive.

Sure, no problem, I could do that.

[Horsepoo!!!]

Originally Posted by OreoCookie

That doesn't invalidate what I've said. All modern filesystems are case-sensitive and all that are in use (with the exception of HFS+ for OS X client users). And in French most of the words are lower-case, so you're not trained to look for the difference.

Oh so now that you know that I'm francophone, I'm "not trained to look for the difference".

I just tested NTFS and while it may have the option of being case-sensitive, I assure you that the default behavior is case-insensitiveness...and for good reasons: reasons that I've already mentioned a few times already.

If they do, they do it on purpose. So no, there is no confusion. Usually you don't create two files with names that differ by case, but if you do, you know why. And that it makes a difference.

But let's pretend for a moment that this behavior was the default and that someone accidently named a file the same as another with the sole distinction of an upper-case letter. You may think this kind of error may happen in 1% of cases...ok...but that's still 1% too many.

Case-sensitiveness is not worth the trouble. *Especially* considering apps are moving towards other means of identification through metadata (other than the file name)...in fact, 'file name' is pretty shitty metadata in the first place considering people rarely name real life objects unless they're very unique.

Do you give a names to your paper document or, say, your bathroom towels to distinguish the differences between them? 'course not. I guess this is a whole other story though.

So all the people who prefer case-preserving filesystems wouldn't feel a difference.

Sure, no problem, I could do that.

Good luck.

[OreoCookie]

Originally Posted by Horsepoo!!!

Oh so now that you know that I'm francophone, I'm "not trained to look for the difference".

Yup, that's what I said. Only a few languages use a lot of upper case.

Originally Posted by Horsepoo!!!

I just tested NTFS and while it may have the option of being case-sensitive, I assure you that the default behavior is case-insensitiveness...and for good reasons: reasons that I've already mentioned a few times already.

Try again. The filesystem is case-sensitive. Higher levels Windows won't let you create 2 files differing by case for compatibility reasons. Read up here, for instance. So you might argue that in the end it doesn't make a difference, and I agree. Still, Apple can do the same thing with ZFS

Originally Posted by Horsepoo!!!

But let's pretend for a moment that this behavior was the default and that someone accidently named a file the same as another with the sole distinction of an upper-case letter. You may think this kind of error may happen in 1% of cases...ok...but that's still 1% too many.

You kinda like this 1% figure, seems to make things really small. I've never heard of anyone trying to create a file where only the cases differ on purpose without knowing why.

Originally Posted by Horsepoo!!!

Case-sensitiveness is not worth the trouble. *Especially* considering apps are moving towards other means of identification through metadata (other than the file name)...in fact, 'file name' is pretty shitty metadata in the first place considering people rarely name real life objects unless they're very unique.

Caring about whether a filesystem is case-sensitive or not isn't worth the hassle. Especially since we'll have more metadata anyway. Byte-comparison is easier to implement on a filesystem-level. Plus you can still have your filesystem to be case-preserving even if the underlying filesystem is case-sensitive.

Originally Posted by Horsepoo!!!

Do you give a names to your paper document or, say, your bathroom towels to distinguish the differences between them? 'course not. I guess this is a whole other story though.

No, my bathroom towels all have different names. They are stitched on the top-right corner so I know exactly which towel I have.

Bottom line is: even if the filesystem is case-sensitive, applications may still emulate case-sensitive behavior. See NTFS if you want an example. Hence Apple could use ZFS and the end-user wouldn't see the difference.

[Horsepoo!!!]

Originally Posted by OreoCookie

Yup, that's what I said. Only a few languages use a lot of upper case.

Try again. The filesystem is case-sensitive. Higher levels Windows won't let you create 2 files differing by case for compatibility reasons. Read up here, for instance. So you might argue that in the end it doesn't make a difference, and I agree.

Whew. That's all I needed to hear.

It's still a bad idea for this case-sensitiveness to exist at the FS level though...especially when OSes make sure users can't create two files with the same name. It just adds more to the confusion.

Explain to me again why case-sensitivity matters?