Updated Zstd Implementation For The Linux Kernel Coming Soon

[phoronix]

Phoronix: Updated Zstd Implementation For The Linux Kernel Coming Soon

While the Linux kernel is increasingly supporting the use of Zstd for various compression purposes, the current Zstd code within the kernel is out-of-date and efforts so far to re-base it against the closer to upstream Zstd state have been stalled. Fortunately, a new attempt at getting the Zstd code updated for the Linux kernel will be published soon...

https://www.phoronix.com/scan.php?page=news_item&px=Zstd-Updated-For-Kernel-Soon

[skeevy420]

Great news.

I hope Zstd fast support is on the to-do list (mainly since BTRFS doesn't have LZ4).

[timofonic]

I hope he succeeds. Waiting to know more about the progress. Thanks for informing us!

[cl333r]

I think people wonder why updating Zstd in the kernel is so complicated.

[evergreen]

I think people wonder why updating Zstd in the kernel is so complicated.

Just follow the LKML mailing list on this topic.
It all comes down to a single asshole Linux maintainer, Christoph Hellwig, who loudly complained about a tiny cosmetic aspect of the patch (something as important as the character case of a prefix of function names), and made it appear as an horrible end-of-the-world life-and-death decision, so much so that most other Linux maintainers, who for the most part don't understand a single word of the topic, just conclude that "if there is smoke, there must be fire", and just stayed clear of the topic.

For the minority who cared and understood the topic, it was an infuriating situation, and it was hard to keep calm. That's a trap, as someone who looks angry immediately loses community respect, so after a few burned people (try to argue with a troll), it felt also safer to stay clear of the topic.

Finally, the main maintainer of the interface (called the "crypto" interface even though it includes data compression), who's the only one in position to act and could express judgment on the merit of a troll's comment, is essentially himself in maintenance mode. Aka, he doesn't do anything, doesn't comment anything, and essentially acts as a roadblock. Most of the time, he just doesn't show up. In rare cases, after enough pressure has built up, he will accept contributions wholesale, without a single comment. So that its bar : it must "look fine" and "highly wanted", and then okay, no need to read, just include that. You can imagine that with a single warning signal coming from some random loud asshole, that's enough of an excuse for him to not move a finger.

And the infuriating part of the story is that the "tiny debatable issue" that the initial asshole loudly complained about was even quickly fixed by the zstd maintainer ! So there is nothing to complain about ! But do you think that the complainer would come back and acknowledge that his topic has been taken care of ? No, let's just inflict maximal damage just to feel powerful, let's act as if nothing happened and people who don't want to know more (which include the gatekeeper) might believe that the "invisible" problem might still be there, and thus stay clear of the topic.

And that's where we are. We, users, are denied the benefit of years of development, even after countless beta testers report flawless experiences and large gains, on the ground that a single asshole wanted to feel powerful enough to mess with someone else contribution on laughingly ridiculous trivia, and it works because the main gatekeeper is essentially lazy, quick to find excuses to not do anything, showing essentially no more incentive and no more interest into his own topic.

[discordian]

I think people wonder why updating Zstd in the kernel is so complicated.

People are wary of big changes and the old zstd code was a one-time custom code drop not following upstream, while the new is automatically exported from the upstream repo. Basically it's no patch, but 100klocs of files replaced with another 100klocs files.

Thats no excuse to merge it, but this raised a few red flags (the only technical reason)

In the past the zstd apparently did not add the right people (zstd compressed kernel), doesn't seem to be the problem now.

And there are the people (I believe maintainers) that just ranted about it being developed by Facebook or something...

[cl333r]

Just follow the LKML mailing list on this topic.
It all comes down to a single asshole Linux maintainer, Christoph Hellwig, (...)

TL;DR: The problem: a troll with issues and a gateway keeper who lost interest because he's prolly been doing it for free.
The beautiful problems of Open Source.

[caligula]

TL;DR: The problem: a troll with issues and a gateway keeper who lost interest because he's prolly been doing it for free.
The beautiful problems of Open Source.

Crypto plugins and transparent file compression also aren't on the top of the list of world's sexiest programming topics. Maybe they don't have enough staff. I've noticed that many MIPS and ARM boards are also lacking hw crypto acceleration. The support may appear later, but only long after the device has already been obsolete for years.

[nadir]

Just follow the LKML mailing list on this topic.
It all comes down to a single asshole Linux maintainer, Christoph Hellwig (...)

I was eagerly waiting for the zstd update and thought I kept up with the mailing list, but I wasn't familiar with how the mailing list worked and was following only the newest thread. I was confused because all the people involved seemed genuinely interested in getting it into the kernel.

Your comment made me search for his name and I found the older threads where he did his thing and wow, what a dick! Gave me a flashback to how his kind dominated certain parts of the internet when I was young.