@30p87 I hesitated briefly to leave all my existing followers and community there, but the downfall made it inevitable and I have really found "home" on Mastodon since. Fewer followers maybe, but with more and better engagement, fewer nazis and all around just a good feeling of being on the right team.
@RandomDamage tagged VLAN is what I proposed originally so the first problem you have to solve is making the NIC emit tagged frames 😏 which is much harder than clever solutions to make TCP connections or something
@RandomDamage ooooh can that be done with something in /sys? is that all the tooling does? that would be fascinating, I'll have to spend some time looking into that
@RandomDamage yeah I don't know why I've always assumed they were something more special than that. Perhaps a long long time ago they were, like 2.4 and early 2.6 kernels
I've seen a few people put content warnings on their "looking for work" posts.
I recommend you DON'T do this. Not everyone who comes across your post will bother to expand it. Worse, content warnings prevent your post from appearing in Trending lists, limiting your reach.
I'm not sure what purpose a CW would even serve here—if someone finds the fact that you're looking for work offensive or traumatizing, that's a them problem, not a you problem.
NVDA has the addition of magnification on their roadmap. I feel like that should be a separate project. Still under NV Access, still sharing some of the code base, and made to work with NVDA, but separate. I don't need and can't use magnification, so I don't love the idea of the download getting bigger to house those features and the startup slowing down as NVDA does visual things that are of no use to me. Am I wrong? No really, I'm asking. nvaccess.org/post/nvda-roadmap…
it made me a bit sad that font support for multiline displays was mentioned as long-term, not even a medium goal. Feels a bit disconnected to me for that reason. If you're going to improve multiline support in "short-term" you're not going to add the font info as a medium-term goal? Gosh oh gosh. Honestly NVDA's Braille support has always been a reason to jump JAWS for me, so if this becomes info JAWS can deliver to me on multiline they can expect me to jump ship more often if not entirely to JAWS. Sad but true.
I dobt it adds a significant size on downloads. I think this will benefit much more than a separate product, namely because the interceptiom between screen reader users and heavy magnifier users is big.
Separate product means new sbraneing, new infra, new release cycle.. for what? To compete with windows magnifier? An integrated approach like Supernova is where the oportunity exists.
Thanks for the thoughts on this, everyone. I didn't realize how many people use both a screen reader and magnification. To me, the two camps have always been separate, with a few people in both. I guess I was wrong there.
I think my concern is that NV Access may lack the staff to maintain/improve NVDA *and* make a magnifier. I also don't want this to be the start of NVDA getting bloated like Jaws is. NVDA is small, simpler, and keeps a lot out of core.
After I left Ai Squared/VFO, I used Windows Magnifier with NVDA full-time. There have been add-ons in the past that make that a little bit easier. But it definitely could use a bit more on board. It has been tough with magnification, still has some bumps at times to get around. Full disclosure, I serve on the windows low vision community panel. So, this is a conversation that does come up from time to time. And splitting off into a separate branch is absolutely positively in no way thinkable bad idea. Well I'm not thrilled with how LP Das/magic/Zoom text was incorporated into jaws, having a single repository that brings things in with modules is much better than doing separates. That was unthinkable 10 years ago, it's common place now
Dodgy take IMO man. Think of how many low vis and losing vis folks this could help. How often are you on a connection where it'll make more than a couple of seconds difference?
@prism Why make newcomers jump through extra hoops to get an add-on when having it in core would only mean a few extra seconds of download and one extra option on the start screen? I'm having a hard time believing anyone's really going to be inconvenienced by that, and am also kinda miffed at such loud reluctance to help more people with more use cases.
@Scott @prism Yeah I see why some might think it shouldn't be included by default, but there's a very easy solution for this. Have a prompt at the start of screen, asking if you need magnification, if so, download whatever relevant modules are needed and go from there. Does not have to be complicated, and everyone gets what they want. Personally I don't really have much of a problem with it being built-in as long as screen reader performance is still good.
Also didn't fully grasp how bad the magnification situation on windows is right now, I've never used it in my life so didn't have a reason to look into it much. Kinda like sighted people learning about screen readers for the first time lol.
@ZBennoui @Scott @prism I didn't realize how bad things are, either. I'm coming around to magnification being included, though I still feel it would make more sense as an add-on or second program meant to work closely with NVDA, something that NVDA can automatically install if desired. My main concern is still bloat and scope creep.
@Scott @prism Ok I know this isn't as related, but as a totally blind guy I've sometimes wondered what magnification software actually does? Like what does it mean to change the size of text? IDK if anyone here can explain but just putting it out there.
@ZBennoui @Scott @prism Think of it like changing how easy it is to feel the lines on tactile graphics. Have you ever felt an image with thin lines, all close together and jumbled? It was hard to track the lines, the image didn't make sense, and it was easy to lose track of what was going on because the lines were too thin or faint. Now give that same tactile image proper lines and more room to breathe. It's much easier to feel.
@ZBennoui @Scott @prism It also gives ability to allow for different forms of low vision. Do you do best with dark text on light background, or the reverse? It allows the user to focus on the part of the screen where the action is, maybe magnify it more or change other attributes. People with normal vision make adjustments automatically, but in that continuum where vision is deteriorating, until the point where it isn't helpful any more, magnification software fills in. this from a totally blind guy.
If so, help us keep up the momentum by sharing with your friends, especially those outside of the typical contributor community. Let’s end the year strong!
Me: I'd like to connect to another account Plaid: Great, which one, and provide credentials Me: Here ya go Plaid: Looks good, they sent an OTP, please provide it Me: Here ya go Plaid: Something went wrong. Try again? Me: Sure Plaid: Great, which account, and provide credentials Me: Here ya go Plaid: Looks good, they sent an OTP, please provide it Me: Here ya go (2nd OTP) Plaid: Something went wrong. Try again? Me: No Plaid: Are you sure you want to cancel and lose your progress?
worst part is that Plaid just breaks your bank's ToS if they never integrated (sounds like they didn't) by scraping their website and logging in with your stored credentials you just gave them.
oh and then they also harvest all of your bank balances/activity/transactions to analyze, just like Zelle does
Aaaaand this would be why I refuse to have a firearm in Suburbia (former), or Hippy Ville (Santa Cruz, current), or in the middle of a big fucking city like San Francisco for "Self Protection".
We shoot sheet rock walls to find out how many walls it takes to stop a typical home defense round. We do these tests on our private, closed gun range. Don't...
@feld I hung some sheetrock recently, and did some insulation so I'mma Guestimate (yes Auto, Guestimate is a word) ...
I'm thinking the hollow points might have flared with some Owens Corning Pink. Would have made zero difference with FMJ/Ball.
Now 2" insulation *Board* would be roughly a 5/8" drywall board-ish, so I'd take 30-35% off of any penetration from a pistol. *Perhaps* 15-20% from shotgun pellets. High energy rifle round, 10-15%?!? May be?
On the occasion of SimpleX and Session getting cryptocoin-donations/funding ... we won't hide that most delta contributors are pretty skeptical of cryptocoin-circles. Exhibit 1: #Webxdc apps - mini apps that anyone can whip up -- and the declared motto still on the webxdc.org web page since inception mid 2022:
No logins, no coins, no platforms ...
and this year, January 23rd, we added "no billionaires" to the growing list of things deserving a "no" :)
How can we evolve and deploy web apps without allowing developers or operators to become a “billionaire in the middle”? There are many ways to try to address this challenge and some involve the tec...
I'm not sure what point this post is making. Delta Chat also accepts bitcoin donations. A bitcoin address is provided on the Delta Chat donation page. Is that a bad thing?
@alandes receiving donations via BTC does not have to do with the chatmail operations (servers and clients) which don't require any chain-coin to operate. Moreover, webxdc was designed as a decentralized web system in difference to "web3" which was typically tied into chain-coin things.
Well, the initial process isn't very user friendly. It's based on web tools for hosting services (it's more like ordering a hosting service than subscribing to a cloud service), not really something designed to be for end users. But once the service is set up, managing the instance is easy.
hi everybody, for those who have read the dungeon crawler Carl series. What do people think of it. I am kind of thinking about getting it. Next time I subscribe to Audible.
MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license. - update README.md maintenance mode · minio/minio@27742d4
Really there's no Split Braille add-on for NVDA? I swear there was one I saw, but it's not in the store, nowhere to be found. Hmm. Braille Extender doesn't appear to do it, unless it's not in the feature list. Loved it in JAWS on a single-line display, now that I'm using 24-by-4, it would be even more useful in things like tables and sometimes I'd use it with my GitHub PR window in one part and my VSCode copy in the other. Just, all round useful, wish I had it for NVDA.
@Jage yeah, and I think the code for the add-on is getting integrated s an official driver inside NVDA 2026.1. iOS already had the support, JAWS also needs a config file to make the multiline work and displaying text across two displays. FS is supposed to address that in their next update. github.com/tactile-eng/cadendu…
Cool. It makes sense for all of this multiline code to converge instead of there being separate drivers for Cadence, Dot, not sure what Monarch or Codex has, did I miss one?
@Jage wonder if Grafiti can do multiline Braille or just graphics? That was about the only other one I had some interest in, but it kind of evaporated after seeing my partner's Orbit display have random faults in connecting and then they lost a key while traveling home. Gosh. Cadence feels a bit more promising with the keys large rectangular rubberized plastic, overall workmanship is just so solid, it just needs more consistent SR support and it'll get there. I know all of these use the HID protocol, and there's a standard HID Braille that Cadence exposes, so JAWS had a driver for it even with 2025, just no multiline. Feels like any new display that comes to the market will be HID, and if there's one advantage it's that it's not a uniquely different API for each display but can be uniformly understood as long as both display and SR speak it.
@Jage I saw a prototype of the second-gen Graphiti at CSUN a few years ago, and it could definitely do both. No idea what the selling hardware looks like.
@sclower OK to be fair I hadn't bothered to look at one more recently. Were the dots still larger than normal, or did it just do braille on a bottom status line?
@Jage @sclower I'm really confused by their page because they say, "2400 Independently Refreshable Pins in a 60 x 40 array" - great (same as DotPad). But: " 40 TrueBraille™ cells offering Orbit’s signage-quality braille and cursor routing keys" so does that mean that unlike DotPad it cannot do mixed Braille and graphics or Braille on the multiline? I am not sure. It's weird they call them out separately like that though, unless it also can do both. I believe the DotPad has a 20-cell separate display like this too, but that you can still do regular Braille on its 300-cell graphics side? Not 100% clear on this either I guess.
# Micron Announces Exit from Crucial Consumer Business
> The AI-driven growth in the data center has led to a surge in demand for memory and storage. Micron has made the difficult decision to exit the Crucial consumer business in order to improve supply and support for our larger, strategic customers in faster-growing segments...
Well this is... probably one of the most chaotic Sound Canvas MIDI compositions I've seen. File name: arm_daue. Internal title: DAUERND ( NUCLEAR GEAR ) Composed by ARM. MIDI: drive.google.com/uc?id=1xNt1LC…. Original LZH: web.archive.org/web/2018100214….
#Framework16 performance issues continued. It took 117 seconds to run git status on the #HardenedBSD ports tree. The screenshots show a VM with 6 vCPU and 32GB vRAM.
For the first couple hours, when the system boots cold, filesystem access is just about as instantaneous as it ought to be. But after around six hours of uptime, the brand spanking new NVMe SSD starts going sloooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooow. Practically unusable when it gets like this.
You can see in the top output that the CPU is mostly stuck in the kernel.
Because of a stupid event from last week, I decided to sequence the first few bars of Phil Collins 'Take Me Home" on my Ableton Move, with emphasis on trying to get the drum pattern and drum sounds as close to the original as possible.
The song uses Roland TR-909 drums for everything except the Cabasa sound.
There is no way I will ever quite get the DX7 e-piano sound right, so I used something that sounds close enough except for the high-end. I couldn't quite get the chorus sound on the e-piano right. It's too stereo, and sounds weird when collapsed to mono.
The Move doesn't really have a gated reverb, required for the clap sound, so I just resampled it with reverb, and faked the gate effect with an envelope curve.
Came out good man! Congrats for not grabbing the first Karioke search result on YouTube and saying you did it (which is what happened last time I heard a remake of this).
@Scott I mean, I could have done that, but it's way more fun wasting a bunch of time trying to get stuff right that nobody but me actually cares about in a limited environment where it's almost impossible to do such things without compromise.
Tonight join @Liz, for Full Circle from 7 P.M. eastern (US time). Their will be a showcase, and in the third hour will be the next episode of Chickenman. So tune in tonight, and see who is hiding in the showcase. HKCRadio.com
Okay, running into some WTF-ery in the realm of network layer operations and security.
So, I have this network. It has some border routers, and inside the border routers are some core routers. I saw a message on a core router (so, one full device behind the outside edge of my network) that said this: > ospf[1027]: %ROUTING-OSPF-4-BADLENGTH : Invalid length 37614 in OSPF packet from x.x.x.x (ID y.y.y.y), HundredGigE0/0/0/0
Neither x.x.x.x nor y.y.y.y are IPs in my network or in my interior routing protocol at all. They are both several AS's away.
OSPF is based on multicast, and I can't easily see a way for it to get several hops inside my network like that. I don't think there's a way to send a unicast packet that will get read as an ospf packet by the destination host. What the heck am I looking at, here? I feel like I am missing something obvious but if so, I don't know what. (which is kind of implicit in "missing," but, you know).
Pro tip: if your Apple Watch is telling you it will take 2 hours to install a software update, just disable bluetooth on the watch. but before you do that make sure the watch itself is connected to your wifi network. that way, the watch will have to use wifi to download the update. worked like a charm when I did it just now.
@Bri sadly this behavior I remember being there even in WatchOS 4.0, nothing has changed since then. Of course back then Bluetooth was even slower on something like the Series 3, so you'd know when it would jump to using it because the update timer would jump from 30 minutes to 4 hours. Even in those early days the Wi-Fi was far faster.
This is a repository of apps to be used with your F-Droid client. Applications in this repository are official binaries built by the original application developers, taken from their resp. repositories (mostly Github, GitLab, Codeberg).
- you tried to install `requirements.txt` 18 times this year. Doing better than last year! - of the packages you installed 67% started with py, 11% python, and 6% Py. You guessed wrong 85 times. - your love for building source has no bounds, except maybe the 92 failed compiles - you updated `requests` 18 times. Urllib is feeling lonely. - the average time between updating `pip` was 97 days. But we warned you 338 times!
Hmm, I'm wondering whether there is a way to use the Eloquence community dictionary with the Android app? No idea what sort of format it requires, does anyone know more?
so, a while ago, I have been tricked into trying out a vibe-coded project. "tricked", because it wasn't disclosed anywhere, and just from screenshots and the readme, it looked fine'ish.
it was a small web-app used to upload and download files. you know, kinda like wetransfer, but self-hosted.
I gave it a quick try, and it worked well enough for the use-case I had (requesting one file from a user). except that.. chunked uploading was broken.
oh, and also: the project provided docker containers and a compose file, but.. the uploaded files weren't stored in the volume at all, they were just stored as part of the ephemeral container filesystem. very much a "you had one job, and you failed at it" kind of situation.
there was a github issue mentioning existing upload issues. I left a comment there pointing out the volatile storage, because I'm not a jerk. also, I felt it's kinda important to note somewhere, because that puts users at risk of actually losing data. that was at the end of october. just after leaving the comment, I noticed an earlier comment by the maintainer. it read:
Thanks a lot for reporting this bug! I actually struggled with the same issue during development. The v3.2.5-beta release was meant to fix it for good, sorry it didn’t work as expected. Just so you know, I’m currently working on v3.3.0-beta, where I’m rewriting all the upload/download logic from scratch.
and then it hit me: the maintainer actually had zero clue. looking at their commit history, it's pretty clear that there's very little - if any - architectural thinking. a lot of the commits were just "give up and rebuild everything", because that's what you get from GenAI: when they struggle writing code, they tend to suggest throwing everything away and starting from scratch.
anyway, I rm -rfed that project, but because I commented in a bug, I was still subscribed. just got a notification from the maintainer posting that a new "beta" is out, so of course, I checked out the commits.
the latest real commit:
[REFACTOR] Change upload and download architecture
oh dear. yes. once again, instead of fixing shit, it's just throwing away everything. and, hold your horses:
205 files changed, +28939 -15962
and sure, lots of that is also docs changes and i18n file changes, but it's literally throwing away everything and "writing it" from scratch. the docker container is now also installing and running minio, because of course that's perfectly reasonable to do. and, even funnier: the "important upgrade notes" - clearly AI generated - mention a lot of things, but what they don't mention is that if you already have uploaded files and you just build a new container... well... you'll no longer have uploaded files.
it's a total feverdream, and one of many similar stories I've seen in GenAI-driven projects. call me a hater all you want, but I, for once, am really happy about stories like that. it's been 2.5 years since people yelled at me with "GenAI will take your job in a few months", and they're still so laughably bad, that - as someone who knows how to debug and actually fix shit - am becoming less worried about my career as time goes on.
what wonderful times we're in.
(also, you can probably google these quotes and find the project and the maintainer. don't. don't harass them or their projects. for all I know, they might have the best intentions and are just heavily misguided by the fearmongering/FOMO-driven/hype-driven marketing of GenAI companies.)
Micron is shutting down Crucial, its longstanding brand for consumer-focused RAM kits and SSDs, as it shifts its attention to supplying memory for AI companies.
the Humble Bundle download experience is fucking terrible I can't even express in words how awful everything about it is.
Oh I have 175 items to download and I can click them all manually or use the "bulk download" button which DOES NOT WORK BECAUSE IT GIVES RANDOM ERRORS FOR SOME OF THE ITEMS AND STOPS DOWNLOADING
ok so I saved the download page HTML and made Qwen write a script to extract all the download links for the format I wanted, ran that through curl, and now I have the files.
I can't believe I had to resort to that, what a bullshit website
In late September, the United Kingdom’s Prime Minister Keir Starmer announced his government’s plans to introduce a new digital ID scheme in the country to take effect before the end of the Parliament (no later than August 2029).
If you find that you're motivated to boost 20 posts in a row from one feed it might be better to merely point people to it and say hey go here it's a worthy follow. 😁🤣🤣
Me cago en todo lo que se menea. El Ayuntamiento tiene funcionando un centro sociocultural o cosa por el estilo en el bajo de mi edificio. Y los miércoles tienen coro. Y no está correctamente insonorizado.
The fact that it was possible for Cambridge UIS to “retire” Talks.cam over the objections of the entire Faculty and replace it with an expensive and unfit-for-purpose enterprise shittified program shows that the Tail is fully wagging the Dog at Cambridge now, and the days of faculty sovereignty are well and truly over.
In a partnership with Mistral AI, CNRS is deploying a locally-hosted LLM chatbot for its employees.
They provide some examples of tasks you can ask the chatbot: summarizing a long text; translating; and writing an outline for an article.
Tasks you'd want to delegate to an assistant because they're beneath you.
And then they decided it would be a great idea to "honor" Emmy Noether, one of the most influential women mathematicians ever, by naming this chatbot Emmy after her.
I cannot fully express how angry that makes me.
Can't be arsed to read something yourself? Ask good old Emmy to summarize it for you, that's right up her alley, she'll do it between proving two major results in algebra or mathematical physics, too bad she's only a website and can't make you a cup of coffee while she's at it.
@cazencott@lipn.info je viens de me rendre compte que le nom de domaine est "alan-poc". Je me demande donc quelle autre figure de l'informatique avait dû être choisie pour nommer l'outil initialement
Lucy
in reply to daniel:// stenberg:// • • •daniel:// stenberg://
in reply to Lucy • • •