Snow Leopard Build 10A286 Seeded to Developers

[Eric S.]

The human application has the most trouble with it - case-preserving file systems are better for wetware.

What is wetware?

Humans aren't the problem, it's applications like backup programs. So for example if you backup and restore two files named TEXT.TXT and text.txt you end up with only one of them (and which one is anybody's guess).

 

[AidenShaw]

What is wetware?

You are.

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


And the example you show illustrates a wetware problem. Wouldn't happen with a case-preserving filesystem - all variations would refer to the same unique file.

 

[loveturtle]

What do you mean by sharing the same amount of total space?

That's how ZFS works.. "Filesystems" (lets call them datasets) are created out of a "pool" of devices (like hard drives). All filesystems share the same total space (the total space of the pool) unless you use quotas or reservations.

Anything other than case sensitive sucks ass btw, but It would be nice on osx to accommodate the few really ****** applications that don't work with it.

 

[Eric S.]

And the example you show illustrates a wetware problem. Wouldn't happen with a case-preserving filesystem - all variations would refer to the same unique file.

I assume you mean case-sensitive (HFS+ is case-preserving). And no, you still have the problem that applications are not written to understand case-sensitivity. In the example I gave, the backup app internally thinks that TEXT.TXT and text.txt are the same file, because it ignores case. It's not a human problem. As people who have tried to use case-sensitive HFSX filesystems have discovered.

Sure, you fix that by fixing the apps. But that's the state of things on the Mac now, where case-sensitive HFSX is available but hard to use because various apps don't play nice. The point is, I assume that ZFS doesn't magically bring a solution for this.

That's how ZFS works.. "Filesystems" (lets call them datasets) are created out of a "pool" of devices (like hard drives). All filesystems share the same total space (the total space of the pool) unless you use quotas or reservations.

OK, I get it. Thanks.

 

[AidenShaw]

I assume you mean case-sensitive (HFS+ is case-preserving). And no, you still have the problem that applications are not written to understand case-sensitivity. In the example I gave, the backup app internally thinks that TEXT.TXT and text.txt are the same file, because it ignores case. It's not a human problem.

In a case-preserving file system, TEXT.TXT and text.txt are the same file, so there is no issue with the backup app. There's only one file, so the correct one is restored.

My definitions would be:

Case-insensitive: User specifies file in any case desired, filesystem upcases (or lowercases) for creates and lookups. User asks to create TEXT.TXT, filesystem creates text.txt. Lookups of Text.TXT, text.TXT, TeXT.TxT all find text.txt.

Case-sensitive: User's spec is stored and looked up exactly. TEXT.TXT and text.txt are two different files.

Case-preserving: User's spec is stored exactly, case blind on lookups. User asks to create Text.Txt, filesystem creates Text.Txt. Lookups for TEXT.TXT and text.txt find Text.Txt. (NTFS)​

 

[Eric S.]

In a case-preserving file system, TEXT.TXT and text.txt are the same file, so there is no issue with the backup app. There's only one file, so the correct one is restored.

That's right; there is no problem with a case-preserving, case-insensitive filesystem (like HFS+). The problem arises when you try to use that app on a case-sensitive filesystem when the app doesn't expect case-sensitivity.

My definitions would be:

Case-insensitive: User specifies file in any case desired, filesystem upcases (or lowercases) for creates and lookups. User asks to create TEXT.TXT, filesystem creates text.txt. Lookups of Text.TXT, text.TXT, TeXT.TxT all find text.txt.

Case-sensitive: User's spec is stored and looked up exactly. TEXT.TXT and text.txt are two different files.

Case-preserving: User's spec is stored exactly, case blind on lookups. User asks to create Text.Txt, filesystem creates Text.Txt. Lookups for TEXT.TXT and text.txt find Text.Txt. (NTFS)​

Yes, I agree.