С обновлением интерфейса Youtube на 24" 1080 мониторе я могу видеть на странице только 6 видео или 3 видео и 5 шортсов. Хотя, зачем вообще показывать шортсы в рекомендациях на десктопе?

Вот бы еще в самом гугле на одном экране было видно только первые 2 результата, одним из которых был бы "обзор от ИИ", вот тогда это было бы UI consistency

Of course it is very sad that the 3 million people of Jamaica have to endure the lethal torrential floods, landslides, & ~300km/h winds of Hurricane Melissa. But surely we can all agree that is the necessary sacrifice we all must make to ensure that the murderous thugs running the world’s fossil fuel industries continue to enlarge their absurdly huge incomes they insist are their right and due.
#ClimateChange #Hurricane #Mellisa #Catastrophe #FossilFuels

Random tip for comparing the performance of Rust and C programs that use stdout:

Rust stdout is line-buffered, even when redirecting to a file.

Glibc (and most C libraries I know of) are smarter than this, and will switch to a more aggressive buffering scheme if they aren't writing to a terminal.

Running the program _under hyperfine_ is enough to trigger this behavior, because hyperfine takes over the stdout stream. Glibc senses that this is not a terminal and turns on a 4096-byte buffer; Rust continues making one syscall per line.

So if a Rust program that prints stuff seems to lose a performance advantage over a C counterpart when run under hyperfine, it's worth stracing to check.

(There's continuing talk of fixing this default in Rust, but it's not fixed yet.)

This entry was edited (22 hours ago)

nlpol

Sensitive content

in reply to Sylvia

nlpol

Sensitive content

Rust's NamedTempFile made me realize that temporary files that delete themselves when the object goes out of scope... are quite nice, actually. You have to structure your code so that they get passed around properly, but that makes the lifetime of the tempfile very clear. It's not garbage to be cleaned up afterwards (if you remember); it's just a temporary resource like any other.

About #Zed accessibility. Of course, it's a sad story. Again, splendors and miseries of #OpenSource. github.com/zed-industries/zed/…

So one thing I notice about #blind Internet culture: even back on Twitter, and now here in the #fediverse, blind people tend towards having discussions in giant threads, sometimes with as many as 10-12 people in them, that can often stretch on for days. I rarely (if ever) see sighted culture do this. I wonder why? It's not a criticism, it's just interesting to me. Maybe because Discord and other chat apps were historically less #accessible, so blind culture tends to use the fediverse more as a discussion platform? Or maybe it's something UI related that makes it easier for blind folks to track giant threads of doom? The few times I've been involved in this style of discussion with sighted folks, they've become confused and begged for everyone to move to Discord or Slack or somewhere. On the other hand, I rarely see blind people do a single, lengthy post broken up and threaded the way sighted people do, with (1/N) at the end. We tend to just move to instances with longer character limits, or put our long form thoughts on a webpage or something.

Edit to add: I'm pleased to say that this post has now become a perfect example of the thing I was talking about; my last post in the thread included the phrase "transsexual furry puppygirls". It makes me happy that people unfamiliar with what I'm talking about need do nothing more than look at the thread on this post.

This entry was edited (18 hours ago)

modulux reshared this.

After 2 years as the first Managing Director of the Foundation, Robin is passing the baton — "I'll see you in the wild blue yonder."

Amandine will serve as acting MD while the Foundation finds the right person to take the org to new heights. matrix.org/blog/2025/10/farewe…

The write-up of my new graph layout algorithm for SpiderMonkey is finally live.

We built a custom layout algorithm for JS and WASM that follows the structure of the source code. No more spaghetti nightmares from Graphviz, and thousands of times faster.

spidermonkey.dev/blog/2025/10/…

in reply to 🅰🅻🅸🅲🅴 (🌈🦄)

gore, death, and a little bit of infodump about h&s, and "fancy" gases used for industrial purposes.

Sensitive content

This entry was edited (21 hours ago)

TIL that the president of Signal believes that people who run Mastodon and/or Matrix servers do so "in most cases" on hyperscaler* infrastructure.

This is my Mastodon server. And its UPS. And its networked KVM for when things get really hairy.

It's also my Matrix server. And Nextcloud. And Git. And Home-Assistant. And Jellyfin. And SearXNG. And Peertube.

When people objected to her claims, she doubled down and proclaimed condescendingly that we "don't have a clear understanding of this space".

TIL that I don't feel confident in recommending people to use Signal. Something's very off here.

*) "hyperscaler" basically means the big cloud infra providers with provisioning APIs that allow you to scale your resources up/down automatically with usage

#Signal #Mastodon #Matrix #SelfHosting

Signal's president claimed it takes billions to replicate the availability and reliability of "hyperscalers" (AWS/Google/Microsoft/Cloudfare) that Signal uses.

#chatmail and #deltachat are about disproving this claim by

1) making relays super cheap (DONE)

2) enabling chat profiles to use multiple relays redundantly (WIP)

3) distributing relay knowledge among chatters (TBD).

Fat servers, corporate overlords and billionaires: not needed and better to not exist for a convivial e2ee future :)

This entry was edited (20 hours ago)
in reply to blazr

@blazr with 2.22.0 you can try out calling, look for "Debug Calls" in the experimental settings :)

Instead of big email providers, we actually recommend self-hosting chatmail.at/relays, or using one of the existing public ones. This is basically how we aim to do scaling, horizontal instead of centralized.

Google is going to make HTTPS required by default in Chrome in a year.

In the post there is quite a bit of talk about the problem of obtaining a cert for local network names. Hopefully their push to make everything-HTTPS will include local network addresses too. We really badly need it.

They kind of seem to say they will, but it's all talk until shown otherwise: "In the future, we hope to work to further reduce barriers to adoption of HTTPS, especially for local network sites."

security.googleblog.com/2025/1…

#chrome #security #selfhosting

"there's no realistic alternative to AWS and the other hyperscalers." (for signal)

except federation. -> #matrix

it.slashdot.org/story/25/10/28…

GNOME 49 (which was shipped in freshly released Fedora 43 and Ubuntu 25.10 and also in rolling distros like openSUSE Tumbleweed and Arch) has a bug with some apps in XWayland, where they'll freeze if you have more than 1 screen.

Details:

gitlab.gnome.org/GNOME/mutter/…

gitlab.gnome.org/GNOME/mutter/…

The only workaround is to turn off all monitors except one. Or, if your app supports it, run with native Wayland instead of XWayland. Or wait for a minor release of Mutter before upgrading to GNOME 49.

in reply to Hubert Figuière

@hub I've found that darktable now works better in native Wayland than XWayland, except for the submenu issue (where it closes the menu when it tries to show a submenu, as it attempts to draw it off screen and hits a bug).

Too bad it defaults to XWayland and has that bug.

(Things like drag and drop into darktable even works in native Wayland, but not in XWayland.)

Fedora Linux 43 is here! What do you get with this update?

* New Anaconda WebUI installer on all desktop editions and spins
* @gnome 49 and being Wayland only
* Fedora CoreOS now based on bootc instead of rpm-ostree
* Under the hood changes like RPM 6.0
* All your favorite versions of Fedora getting up to date!

Happy upgrading!

➡️ fedoramagazine.org/announcing-…

#Fedora #Linux #OpenSource

This entry was edited (1 day ago)
in reply to Fedora Project

What's new in Fedora KDE Plasma Desktop 43?

* New Anaconda WebUI installer
* Automatic updates for Kinoite (which is based on Fedora KDE)
* Latest updates from @kde Plasma 6.4.5 with improvements in tiling, color management, accessibility, and more!

➡️ fedoramagazine.org/whats-new-i…

#Fedora #FedoraKDE #KDE #Linux #OpenSource

Any #ZFS folks want to offer me an opinion on the following? I have 2 ZFS pool of raidz2 with 7 2TB SAS drives each (about 11.5T). I want to add a LOG drive. I have a single SAS SSD that's 300G. When I suggested using a 1TB LOG drive earlier, someone said that was way too much. (which is why I picked up a used 300G SSD)

My question is whether it makes sense (is possible) to partition that drive into 2 partitions and have one pool use 1 partition as a log drive (e.g., 150G) and the other pool use the other partition as a log drive on the same physical device.

Is this going to be worse because it's just too much IO on one device? Is it reasonable? Any other ideas?

Thanks
#homelab #linux #selfhost

in reply to Paco Hope wishes ill for JK Rowling

oh gosh, unless you know you are doing a lot of synchronous writes (almost never; databases and NFS mostly... maybe VMs???) you do not need a LOG at all.

If you do add a LOG, it just needs to be large enough to buffer the writes until they're flushed to the other vdevs. You honestly could get away with something as small as like 8GB for a low IO / homelab type environment. But you won't find a modern drive that small, so if you're going to do this just make it as small as possible. Pick quality over quantity here. SLC drives (do they even exist anymore?) are of course ideal.

Splitting the LOG device between multiple zpools isn't crazy. I have a tiny partition I use for LOG and then use the rest for L2 ARC which is another option to maximize the usage of your fast storage.

This entry was edited (22 hours ago)