NVK Vulkan Performance Improving With Mesa 26.2 Against NVIDIA Proprietary Driver

The performance of the Nouveau kernel driver with the Mesa NVK Vulkan driver continues to improve each feature release in its quest toward better competing with NVIDIA's official Linux graphics driver with its proprietary Vulkan driver...
https://www.phoronix.com/news/NVK-Mesa-26.2-Performance

@thomholwerda How come I have like 50-ish firefox tabs open and another 20-ish librewolf tabs open and I can still use jitsi just fine? 8 GB of RAM, Alpine Linux, sway...

I'd say it is a Windows problem too, nowadays windows happily takes 2-3 GB of RAM on idle.

And they call it ‘progress’

I call it unnecessary destruction

When you allow AI to write your alt text.

@Andres4NY @TheEvilSkeleton well it is on mint for not redirecting the user to the correct bug tracker (theirs, downstream)

@TheEvilSkeleton There's this blog post also which I think quite resonates with how I think about this whole situation: https://pointieststick.com/2026/07/19/whos-responsible-for-bug-reports-on-old-software-versions/

@TheEvilSkeleton so I got linked to this before ... from what I can tell in the issue, as an outsider, it looks like neither side wanted a productive conversation, and it just felt rude on both sides. The issue body itself also initially didn't look completely clear what the actual problem is, it just asked for rebranding everything without any specific reason.
This article itself is just inflammatory, again pointing fingers at LM which is not helpful.

Note linuxmint is not innocent either, they definitely messed up by packaging gnome-calendar in this state, but I feel if the discussion started by clearly stating the problem (people reporting long solved bugs in the wrong place, which is hurting upstream) and not pointing fingers all the time the outcome probably would've changed significantly.

But instead there was a lot of miscommunication, both from your side and their side, miscommunication that could have definitely been avoided, which is sad to see.

@me @davidrevoy yep, reminds me of this gem

@davidrevoy Meta AI detects that human art is made by AI... so funny

For Instagram, my art is "Likely made with AI". Hard to be an artist in 2026...

@hexaheximal no need to overthink them

we have macbook at home

Today the last patent for MPEG 4 Part 2 Visual finally expired.

https://meta.wikimedia.org/wiki/Have_the_patents_for_MPEG-4_Visual_expired_yet%3F

#NoSoftwarePatents #OpenCodecs #MPEG4

Hey, everybody? I get that it's frustrating when Codeberg is down. But please remember that it's a community project, and that the expectation of "everything is up 100% of the time" was created by exactly those tech companies that were killing off communities for the past 20 years. 100% uptime was never true before Facebook/Google/etc. came around, not even close.

And it's even more difficult to keep things up nowadays; there is *so much more* malicious traffic coming from the internet now than back then, especially since the rise of LLM scrapers.

So if you want Codeberg to be down less often, might I suggest reaching out to them about volunteering instead? Not right now, because they're probably pretty busy getting things back up, but once things have been stabilized a bit. Then you yourself can help improve the uptime!

#Codeberg

Who’s responsible for bug reports on old software versions?

pointieststick.com/2026/07/19/whos-responsible-for-bug-reports-on-old-software-versions

Consider this hypothetical: An operating system (OS) ships version 1.5 of a piece of software. Meanwhile, the latest version of that software is 3.0. A user on that OS experiences an issue in version 1.5, or has an idea for a new feature. Who should they…

@domi This reminds me of Microsoft servers trying to scrape my website by trying to lookup weird urls: https://berkeley.edu.pl/notice/B3TdAoZFIHjCJa2dTE

so i look into why my logs are so enormous. i shouldn’t have left them as verbose for 1.5 months but they also shouldn’t have grown to EIGHT GIGABYTES. and most of my recent log lines are something like

Jul 04 13:36:22 msg="Limit of simultaneous TCP connections reached - raise max-tcp-connections" prio="Warning"

this, but 15 times a second… Which usually means that we’re under attack. but it’s a very lousy attack, i can’t even see it on the graph, so they’re only really causing some disruption to TCP connections, not UDP. and most of DNS is UDP, that’s why I’m only noticing right now.

anyways. ss -tulnap | grep 53 shows a lot of connections with addresses that begin with 2a03:2880. | wc -l says 273, from maybe 5-6 different hosts. and, my dear comrades, can you guess who owns 2a03:2880::/32?

inet6num:       2a03:2880::/40
netname:        PRN
country:        US
admin-c:        RD4299-RIPE
tech-c:         RD4299-RIPE
status:         ASSIGNED
mnt-by:         fb-neteng
mnt-by:         facebook-neteng
created:        2020-03-11T00:33:38Z
last-modified:  2020-03-11T00:33:38Z
source:         RIPE

that’s right! facebook! and after a preliminary check, we’ve seen this same exact thing on at least one other #SERVFAIL nameserver

in “completely unrelated news”: i “slipped” in my kitchen and executed ip6tables -I INPUT -s 2a03:2880::/32 -j DROP as root on sakamoto i’m so sorry (not sorry). and through the power of ip6tables… suddenly my TCP limits aren’t reached anymore. CURIOUS

if you didn’t have enough reasons to hate meta: FUCK META

@kde @kde Does the TV run kde plasma bigscreen? :P