Google Is Exploring Potentially Using Btrfs In Android

[doublez13]

Why dont they use ZFS instead? It works on 512 MB systems such as Raspberry Pie:

https://github.com/zfsonlinux/zfs/issues/5417

They can't legally "distribute" a Linux kernel with ZFS baked in. Non-compatible licenses. That'd be cool though!

[profoundWHALE]

They can't legally "distribute" a Linux kernel with ZFS baked in. Non-compatible licenses. That'd be cool though!

Uh, yes you can.

[nomadewolf]

Impossible to say, if anything there is a lie, cause there is only the claim that it is wrong, but no explanation what is wrong (in detail) nor why this is wrong. Again I want to know what design mistakes are baked into the odf. I only read repetition of the claim. Bigger codebase than XFS? No surprise, does XFS has volume management? Far too long to stabilize? Did anybody compared how long XFS took to stabilize? Yeah I know, it was so stable from the first day, there wasn't even the need for some repair tool...

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID-56-Is-Bad

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID5-RAID6-Fixed

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID56-Not-All-Clear

https://lwn.net/Articles/393148/

https://lwn.net/Articles/393144/

https://www.google.com/search?q=btrfs+broken+design&oq=btrfs+broken+design&aqs=chrome..69i57j0j69i61l2.8046j0j7&sourceid=chrome&ie=UTF-8

Nothing is perfect from the beginning. And nothing is perfect, even after years/decades. But somethings start in the wrong direction...

[smitty3268]

Uh, yes you can.

It's at best legally questionable. Some people say you can and some say you can't. Probably don't want to bet the farm on that if there's another option that's mostly just as good or could be made just as good with a little work.

Especially when you're google and everyone is just looking for an excuse to sue you anyway. The more money you have, the more lawsuits you face whether they're valid or not.

[PuckPoltergeist]

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID-56-Is-Bad

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID5-RAID6-Fixed

http://www.phoronix.com/scan.php?page=news_item&px=Btrfs-RAID56-Not-All-Clear

This is all not about broken design, especially not odf. It's about bugs in the implementation that got addressed (mostly by Qu Wenruo).

https://lwn.net/Articles/393148/

https://lwn.net/Articles/393144/

I've read the thread again, and Edward was wrong here. Yes he found a bug in the merging code. But there wasn't a fundamental design issue, and again not with the odf.

Nothing is perfect from the beginning. And nothing is perfect, even after years/decades. But somethings start in the wrong direction...

Facts?

[chithanh]

They can't legally "distribute" a Linux kernel with ZFS baked in. Non-compatible licenses. That'd be cool though!

Uh, yes you can.

It's at best legally questionable.

It may be legally questionable, but it is definitely not more questionable than shipping proprietary drivers in the kernel. Which is what most Android phones do.

So a manufacturer deciding to ship ZFS will run no risk in addition to those that they already accept.

I've read the thread again, and Edward was wrong here. Yes he found a bug in the merging code. But there wasn't a fundamental design issue, and again not with the odf.

I haven't seen where Edward Shishkin admitted that he was wrong. Instead he insisted in later posts that the proposed solutions are only workarounds that do not fix the core problem.

In fact, the free space issues that still plagued btrfs afterwards make it appear to me that the underlying problem was not fixed, only worked around so that Edward Shishkin's testcases would no longer trigger the problem.

[PuckPoltergeist]

I haven't seen where Edward Shishkin admitted that he was wrong. Instead he insisted in later posts that the proposed solutions are only workarounds that do not fix the core problem.

And Chris explained the details of btrfs implementation, where he thinks Edward was wrong. In the end in sounds very much like Edward was concentrating on tail packing, what btrfs doesn't do. So Edward had critics and questions, Chris answered and explained and at some point there wasn't any more response from Edward.
The free space problems came from bad/incomplete implementation in the beginning. But that was known. Because of this, btrfs had the artificial limitation of keeping 20% disk space unallocated in the beginning. That's a problem of COW, not a design bug.

[ssokolow]

It's at best legally questionable. Some people say you can and some say you can't. Probably don't want to bet the farm on that if there's another option that's mostly just as good or could be made just as good with a little work.

Especially when you're google and everyone is just looking for an excuse to sue you anyway. The more money you have, the more lawsuits you face whether they're valid or not.

More specifically, here's the relevant bit of section 6 of the GPL 2.0: "You may not impose any further restrictions on the recipients' exercise of the rights granted herein."

That's why two copyleft licenses (eg. GPL and CDDL) have to explicitly allow each other. The copyleft for a license other than your own qualifies as a "further restriction".

The CDDL, on the other hand, inherited the GPL-incompatibility of the MPL 1.1, which it was based on. (Mozilla themselves always released their code under "your choice of GPL, LGPL, or MPL" until the MPL 2.0 fixed that... but the CDDL, perfectly reasonably, didn't keep the MPL's GPL-like "you can obey a newer version of the MPL instead" clause.)

[profoundWHALE]

Look, if you can ship a kernel with proprietary drivers or have something like Android, of course you can have ZFS on Linux. It's as simple as that.

[fuzz]

Look, if you can ship a kernel with proprietary drivers or have something like Android, of course you can have ZFS on Linux. It's as simple as that.

I highly doubt Google is going to touch anything with Oracle's name on it (and I don't blame them).