Oracle Might Be Canning Solaris

[ldo17]

Bandying about the term “zealots” is one thing, pointing out that btrfs has features that ZFS does not is another...

[DanL]

btrfs has features that ZFS does not

Did anyone say it didn't?

[janos666]

wishful thinking they'd re-license ZFS to a GPL-compatible license and invest more in Oracle Linux

Yeah, but... why should we really care at this point, though?
We could already guess Oracle will likely never fund another big ZFS development projects and nobody else will pay for the infamous Block Pointer Rewrite feature (pre-req for actual features like de-fragmentation, RAID profile and/or stripe width change -> things which are already possible in Btrfs, even if Btrfs isn't perfect either, especially when it comes to RAID), so in my opinion it's just another "confirmation" that ZFS will never be "fully developed". -> Well, it depends on the individuals idea about feature completeness but in my opinion Block Pointer Rewrite is something which should have been implemented in the first public release (preferably) or worked on soon after at relatively high priority (and finished in a couple of years at top, not letting it sink away by another "exotic" features making it harder, if not practically impossible to implement around) in order for ZFS to really earn that "The Last Word in filesystems" (because I certainly crave those features hard enough to use Btfs instead).
ZFS can't be my last filesystem if I have to regularly destroy it whenever I want to make a pool bigger and/or faster by adding new disks or simply in order to cure the fragmentation.

[ldo17]

Did anyone say it didn't?

Well, there was

ZFS is miles ahead of btrfs, it's not even a question to debate.

[gigaplex]

so
1) i gave you article by zfs engineer which explains that zfs is obsolete
2) you said it is 7 year old
3) i said then it was obsolete 7 years ago
4) you said cool story
ok, it seems appropriate to say that you are idiot

The article also claimed that btrfs would be the default file system pretty much everywhere within 2 years. Btrfs still isn't production ready 7 years after that article. Especially if you're looking for RAID5/6 parity style arrays. It has the potential to obsolete ZFS, but it hasn't happened yet.

[boxie]

To be honest here, I would love to btrfs to become default everyone once it has sorted out all of its problems with Multi Disk arrays. I think it could be rather wonderful. maybe 2017 will be the year of btrfs by default along with the year of the linux desktop!

[janos666]

The article also claimed that btrfs would be the default file system pretty much everywhere within 2 years. Btrfs still isn't production ready 7 years after that article. Especially if you're looking for RAID5/6 parity style arrays. It has the potential to obsolete ZFS, but it hasn't happened yet.

Well, in my opinion it did. I think the engineers at Sun made a huge mistake with how they handled the infamous Block Pointer Rewrite feature (which would have allowed defrag and RAID profile and stripe width conversions). Not only they decided to release the first public version without it (*) but they kept pushing it further away on their timeline until 1: it's absence became generally accepted, and 2: the later exotic features made it harder (if not practically impossible) to implement on top. And now certainly no company will pay for it and nobody will do it for free (actually, "nobody is able to" might be closer to the truth).

* I can't accept ZFS as "The Last Word" in filesystems. It can't be my "last" filesystem if I have to regularly destroy and recreate my pools (and then restore everything) every time I 1: wish to defragment it, or 2: change disk/RAID the topology/profile. The lack of this feature means the manual destruction and recreation+restoration of pools/filesystems effectively becomes a regularly scheduled routine maintenance job. And I think this is a huge pain for small scale / home users (like me).

[starshipeleven]

Not that simple. Even Linus has said that if a driver was implemented first for another OS and then ported to Linux it may not be a derivative work.

The linux version of that driver or the part of the driver dealing with the linux kernel would. NVIDIA is using a shim recompiled on the consumer's machine to connect their blob to the linux kernel for a reason.

[k1e0x]

Yeah, but... why should we really care at this point, though?
We could already guess Oracle will likely never fund another big ZFS development projects and nobody else will pay for the infamous Block Pointer Rewrite feature (pre-req for actual features like de-fragmentation, RAID profile and/or stripe width change -> things which are already possible in Btrfs, even if Btrfs isn't perfect either, especially when it comes to RAID), so in my opinion it's just another "confirmation" that ZFS will never be "fully developed". -> Well, it depends on the individuals idea about feature completeness but in my opinion Block Pointer Rewrite is something which should have been implemented in the first public release (preferably) or worked on soon after at relatively high priority (and finished in a couple of years at top, not letting it sink away by another "exotic" features making it harder, if not practically impossible to implement around) in order for ZFS to really earn that "The Last Word in filesystems" (because I certainly crave those features hard enough to use Btfs instead).
ZFS can't be my last filesystem if I have to regularly destroy it whenever I want to make a pool bigger and/or faster by adding new disks or simply in order to cure the fragmentation.

Oracle has never funded or done ANY development of ZFS. (meaning the free open source OpenZFS version). All of that was done by the Illumos developers and former Sun developers who wrote it. All Sun development ceased at pool version 28. A re-licence won't happen due to the cold war between Illumos and Oracle over the licence.

[k1e0x]

The linux version of that driver or the part of the driver dealing with the linux kernel would. NVIDIA is using a shim recompiled on the consumer's machine to connect their blob to the linux kernel for a reason.

Nobody knows.. but I don't think it even matters. I think Linux is on it's way out the door, things like this are killing it. Other FOSS OS's don't have this problem.