Tux3 Gets Harshly Criticized Over Code Quality

[stiiixy]

.....

Guess the pre-code review made him cringe enough that he felt that Tux3 didn't warrant a code review, just lots more development.

Besides, that's all semantics. You all get the drift.

And Mark45, you're referring to 'his'. Male person, ownership.

[nils_]

Looks like Philips has replied with some snark of his own. I think Chinner has many valid points from a cursory view of the code.

http://lkml.iu.edu/hypermail/linux/kernel/1405.2/02219.html

[chinoto]

hease is not a word.

Even if it is a word, the first definition I found is incompatible with what he said: Hease@Urban Dictionary.

[danielbot]

Did you read the entire message? There's big fundamental issues, really big. "Some code" is not nearly enough, significant development is required.

It is only fair to point out that Dave Chinner is lead developer for the XFS filesystem, not exactly an impartial observer.

[Gusar]

It is only fair to point out that Dave Chinner is lead developer for the XFS filesystem, not exactly an impartial observer.

Are you saying his analysis of the current tux3 state is invalid? I would say it's exactly the opposite, it's because he is a filesystem developer that he knows what's required for a filesystem to be in the kernel. Who else is going to be reviewing tux3 but other filesystem developers, someone who has no experience with filesystems?

[danielbot]

Are you saying his analysis of the current tux3 state is invalid? I would say it's exactly the opposite, it's because he is a filesystem developer that he knows what's required for a filesystem to be in the kernel. Who else is going to be reviewing tux3 but other filesystem developers, someone who has no experience with filesystems?

We appreciate Dave's review, but the tone of it falls short of collegial or professional.

[nils_]

We appreciate Dave's review, but the tone of it falls short of collegial or professional.

Isn't this to be expected on the LKML?

[danielbot]

Or Btrfs given how he said Tux3 takes largely same design decisions as Btrfs but is years behind.

Tux3 is very different from Btrfs, I don't know where you got that from. Btrfs uses a "shared tree" design similar to Tux2 from back in 1998 or WAFL from even earlier, with the added wrinkle of per-block reference counters. Tux3 has a much simpler single tree design like would you would get if you turned Ext4's inode table into a copy-on-write tree.

[danielbot]

Isn't this to be expected on the LKML?

Sadly, yes, but expected is not the same as admirable.

[TeamBlackFox]

Tux3 looks better than ext4, but I think I'll still with JFS and XFS for the last few machines I have that run Linux. I agree though that Tux3, like BTRFS is not ready for production use. Of course ZFS is the best filesystem I have seen but its not available in the mainline kernels, only on Solaris, illumos and BSD.