Linux 6.7 Released With Bcachefs, Intel Meteor Lake In Good Shape & Nouveau GSP Support

[PuckPoltergeist]

the old unstable code is available for testing, as with bcachefs the old unstable code is available for testing (but behind experimental flag, so more honest)

But why testing something that is known incomplete and will be replaced soon?

[PuckPoltergeist]

What is the question about specifically?

What experimental features were merged without clearly denoting as such?

[varikonniemi]

But why testing something that is known incomplete and will be replaced soon?

the implementation will be replaced, not the interface. So you can test a setup with mock data to be able to immediately deploy it with real data once the stable implementation is released.

[spicfoo]

What experimental features were merged without clearly denoting as such?

You have already the context seeing what I was replying to.

https://www.phoronix.com/news/Btrfs-Warning-RAID5-RAID6

When this was merged, there wasn't anything to discourage it's usage. That was added later.

[PuckPoltergeist]

You have already the context seeing what I was replying to.

https://www.phoronix.com/news/Btrfs-Warning-RAID5-RAID6

When this was merged, there wasn't anything to discourage it's usage. That was added later.

I don't have the complete history, but I'm sure raid56 was flagged experimental all the time. The "Strongly Discourage" was added later, cause people ignored the experimental and complained about data loss.

[waxhead]

You have already the context seeing what I was replying to.

https://www.phoronix.com/news/Btrfs-Warning-RAID5-RAID6

When this was merged, there wasn't anything to discourage it's usage. That was added later.

Incorrect. The manual and wiki said otherwise. This was added to make it more obvious for people that do not pay attention to important things like reading the manual... People like you.... Because it is even explained in the article you liked to.

But yes, this should have been there from day one "raid"5/6 was 'implemented'

[PuckPoltergeist]

the implementation will be replaced, not the interface. So you can test a setup with mock data to be able to immediately deploy it with real data once the stable implementation is released.

Ah okay, I was irritated by your comparison with bcachefs where I assume raid56 as stable. Comparing an unstable/incomplete implementation with a stable one didn't make much sense to me.

[spicfoo]

I don't have the complete history, but I'm sure raid56 was flagged experimental all the time. The "Strongly Discourage" was added later, cause people ignored the experimental and complained about data loss.

I don't think so. You could go back and look at it, I don't recall anything marking it clearly as experimental when it was merged. If users weren't clearly aware of this, you have failed to communicate it correctly as a project. You need to do this in a) the commit that merges this feature b) In the associated user facing docs like the man pages c) in the tooling that enables this feature d) in the project pages and then you need to coordinate all this was distros properly. They did fix these failures later but there was a gap in between which caused a bunch of reputational damage they could have avoided. Btrfs developers I suspect would readily admit they screwed up here before.

[spicfoo]

Incorrect. The manual and wiki said otherwise. This was added to make it more obvious for people that do not pay attention to important things like reading the manual... People like you.... Because it is even explained in the article you liked to.

But yes, this should have been there from day one "raid"5/6 was 'implemented'

I don't use Btrfs and I don't use RAID so zero need to personalize this but it wasn't in the manual when it was originally merged. It was added to the wiki at a later point but it's irresponsible to blame users who aren't watching wiki pages which thankfully Btrfs developers (unlike the fanboys) didn't do. It must be have added to the tooling on day one. This oversight was corrected later which is a good thing. It is a lesson that BCachefs developers should learn from if they ever add any experimental features which are better avoided altogether.

[varikonniemi]

I don't think so. You could go back and look at it, I don't recall anything marking it clearly as experimental when it was merged. If users weren't clearly aware of this, you have failed to communicate it correctly as a project. You need to do this in a) the commit that merges this feature b) In the associated user facing docs like the man pages c) in the tooling that enables this feature d) in the project pages and then you need to coordinate all this was distros properly. They did fix these failures later but there was a gap in between which caused a bunch of reputational damage they could have avoided. Btrfs developers I suspect would readily admit they screwed up here before.

Indeed. Now compare this to bcachefs developer. about 10 months or so ago he said that he is adamant on shipping bcachefs with full raid support. But then raid56 was found to have some corner cases, so it shipped with raid support behind compile time experimental flag.