Slashdot Mirror
ext3fs in Linus' Kernel Tree
peloy writes: "According to Linus' changelog for Linux 2.4.15pre2, the long waited ext3fs, the sucessor of ext2 with jounaling capabilities, has finally made its way into the official kernel tree. I have never tried ext3fs but it looks that now that it is "blessed" by Linus I'll be upgrading my old and trusty ext2fs partitions soon."
Why Linus chose ext3 over the more proven technology, XFS? I have been using XFS in linux on my laptop, and it is rock solid, stable, and fast. It's a proven filesystem.
I expect ext3 to have major bugs once people start using it in production. People will be really pissed when their data becomes corrupted/deleted and could cause even more instability in the kernel.
On another, only slightly related note, I hate the fact that ALSA isn't standard in the kernel - instead they use OSS.. which releases only half-assed free drivers, then expects you to pay for the good drivers.
But I guess when you're the dictator of an operating system you can do with it whatever you like.
So the Linus has said it, so it shall be done, as is the will of the Linus.
I have nothing against the ext3 filesystem or the new virtual memory patch but Linus needs to stop adding these relatively radical changes into the so called stable kernel reserved only for drivers and bug fixes. The issue is not that big for most hackers reading this but alot of us run Linux on mission critical servers that we bet our jobs on. Even before the radical kernel patches, all the newer kernels had big-time showstopper bugs. Many admins even run the old 2.2 kernel to avoid unnecessary problems. I have been running 2.4 and had no problems yet luckily. However I really do not know if I can recommend Linux as a server OS anymore. I want stability and Freebsd and Solaris seems to meet my needs alot more. Hopefully this madness will stop soon. We all love to bash Microsoft for releasing buggy service packs for NT that have not been tested thoroughly but there seems to be no real testing with Linux kernel patches. Freebsd conquered this problem by having 2 development teams. One for the stable branch and one for the development branch. No radical changes are allowed in the stable branch and the stable branch must go under lots of testing to be approved to be released as stable. I now know why BSD hackers love there development model. Cutting edge is great for some users but please do not include it in the kernel where a lot of people count on it for servers and mission critical applications.
http://saveie6.com/
f*ck fsck. Thank Linus it's gone /usr/bin shortly after my new RH7.2 upgrade, even though I had the proper permissions. It turned out not to be some weird ext3 issue, but the fact that the RH7.2 install had secretly used (without my knowledge or permission) two of the dirtiest words in the fs language: chattr and immutable. bah. Having eventually figured out the problem, I now stand firm behind ext3.
Of course, 2 days ago I (and a number of Solaris SA's I called in for moral support) blamed ext3 for some really weird problems with me not being able to write/chmod/touch anything in
ummm... actually if you would read up on windows 2000 you would see that it was based on the win9x clones. But then again you wouldn't be concerned with FACTS now would you
Are there any disk array controllers or other storage devices that do FS-independant journaling at the block level?
It'd be an interesting way of doing backups -- instead of restoring from tape, you could get the disk controller to back out changes to a specific timestamp.
I'm sure there are some gotchas -- without knowledge of the filesystem, a specific hardware level transaction may not have complete filesystem level coherency, for one. And it may take a lot of disk to keep a log of any decent duration.
But it still seems like an interesting way to accomplish some kind of fault tolerance for any OS.