Nerd-snipping of the day - a sharp programmer on a Chinese forum noticed an extremely obscure fact about K&R - its Courier font is unusual and doesn't seem to match any of the modern or common versions of Courier we know today. So now I find myself hunting for technical references of that Autologic APS-5 machine used by K&R. #retrocomputing #unix
The Pastor entered his donkey in a race and it won.
The Pastor was so pleased with the donkey that he entered it in the race again and it won again.
The local paper read: PASTOR'S ASS OUT FRONT.
The Bishop was so upset with this kind of publicity that he ordered the Pastor not to enter the donkey in another race. The next day the local paper headline read:
BISHOP SCRATCHES PASTOR'S ASS.
This was too much for the Bishop so he ordered the Pastor to get rid of the donkey. The Pastor decided to give it to a Nun in a nearby convent. The local paper, hearing of the news, posted the following headline the next day:
NUN HAS BEST ASS IN TOWN.
The Bishop fainted. He informed the Nun that she would have to get rid of the donkey so she sold it to a farm for $10. The next day the paper read:
NUN SELLS ASS FOR $10
This was too much for the Bishop so he ordered the Nun to buy back the donkey and lead it to the plains where it could run wild. The next day the headlines read:
NUN ANNOUNCES HER ASS IS WILD AND FREE.
The Bishop was buried the next day.
So stop worrying about everyone else's ass and just cover your own!
Imagine a free, point-to-point file transfer service with no file size or type limitations. Enter Blip – it's like platform-agnostic AirDrop from anywhere.
The Problem to be Solved When Bart and I record an installment of Programming By Stealth or Security Bits, we both record both sides of our conversation. We use Audio Hijack to record the files in stereo, so he’s on one track, and I’m on the other.
Essentially, yes, just with a UI and without compressing folders into a zip file first.
It's also interesting how they emphasize that the files are transferred directly from point A to point B, but they also mention that they use relay servers. I'm not sure I'd personally trust closed source encryption, even though they say it exists, but that's just a personal thing.
@erion WebRTC requires relay servers, doesn't it? At least for initial signaling, and in case NAT fails. Of course, there's always just wormhole.app where you can trust their encryption, but they have a cloud too.
I worked in an offie. A man came in and said he knew where I lived. I didn't recognise him. He said he lived next door. I said a bald bloke with glasses does, not you. I'd just bubbled his wife's affair, without realising. They split up, sold up and moved on. Eek.
Note for people booking rooms for @nationsblind at the Sheraton. The number on the NFB site goes to agents at reservations.com who have no clue or access about the group rate. I called the Sherraton direct at 504-525-2500, pressed 1 for reservations, then after a couple minutes of voice prompts, got to an agent.
@fox Well, you didn't just read my new graphic fanfic where Harry Potter is impregnated by Pixelfed and gives birth to loops. Because I haven't written it yet.
@fox Also, I am a blind screen reader user. On a more serious note, thanks for remembering to use alt text, even multiple posts deep in a silly thread.
Go back in time and explain to your grandparents that in 2025 the UNITED STATES would be sending its top officials to Europe in support of resurgent Nazi movements aiming to overthrow their democracies.
Has anyone thought about making a #virtual #Braille #display for #Windows that would hijack the touch screen for use with #NVDASR or #JAWS as a Braille input device, and allow using things like a mode for navigation with dot taps like command mode with #BrailleScreenInput on #iOS with #VoiceOver? I'd pay for this. #Accessibility.
More seriously, it would probably need to be integrated into core to work well. NVDA's touch support has all the fundamentals to make this possible, but I think it would be difficult to do it well in an add-on without overriding all of NVDA's other touch commands or without many many hacks.
To clarify, do you mean a virtual Braille display or a virtual Braille keyboard? It sounds like the latter. Maybe something that emulates a device that's solely a Braille keyboard, like an Orbit Writer?
@matt I mean a Braille keyboard, I said display because that's how devices like the Orbit Writer connect, as a display rather than as a keyboard. Now I see how that could be confusing wording!
So this begs the question, why spend money on touchscreen-based Braille input rather than a Braille keyboard? I understand why Braille screen input makes sense on a phone, because of portability. Are you wanting to use that on a Windows tablet?
Not saying you shouldn't want this, just trying to clarify the rationale and see what kind of business case there might be.
@matt I already have my surface pro, and without the keyboard it is basically pointless. Braille screen input on Windows would make windows on touch actually tollorable. The way NVDA currently handles it makes it harder to use than it needs to be.
So it sounds like the cleanest way to do this wouldn't be actually emulating a Braille device, but doing an NVDA add-on that does Braille screen input.
Re: last boost (librepunk.club/@raeaw/11399332…), OK, I admit I'm a little scared about my future employability, particularly in light of the current enthusiasm (even among many developers) for using LLMs to generate code, independent of how well it actually works. It does make me wonder if I should jump into that myself, if only to prove that I'm not becoming set in my ways and at risk of being left behind in my middle age.
IMO you've got the kind of niche expertise (re: the inner workings of screen readers, OS-level accessibility plumbing, etc) that you'll probably be fine employability-wise.
Even if "AI" was as effective as companies thought (which, they'll learn the hard way that it isn't), and the job market does shrink, it'll shrink in the most common sectors, like web dev. There's simply more data there to train AIs from
The real challenge is going to be funding going to fads instead of impactful work
If you're at risk of being left behind in your middle age, I'm at risk of being left behind in my young age. Fuck using AI's for coding, they have absolutely no context and just fall over when you ask them to do anything more complex than "generate this react view" or "write me a python function to do x". Sadly, that's all a lot of developers do anyways, so having a tool do it for them isn't a very big strech.
the hate for America from Canadians is justified but also not fun to see because now I'm seeing people completely wanting to stop selling stuff to the US which means if I buy Canadian, I'm paying Canadian prices which is fine but also, as a person living in the US, I feel like we're being attacked.
Today's social media irritation? People who boost, repost etc the 'on this day' things, late. Double-dick points when they mention the date so your deliberately doing it wrong.
Honestly #riffusion is making me wish I was actually musical, and knew how to work with stems and remix things. All the examples I've posted here are just silly fun, but unlike with Suno or Udio, I can regularly describe things to Riffusion in ways that get it to generate music that I actually enjoy. If I could, like, trim the start and end, and remove part of one chorus, and move the verses around a bit, some of these would be perfect. It somehow feels worse to get a song that's almost exactly what you imagined it would be, but just not quite, than to think of a fun idea for a song, but know you can't play or sing or edit, so it's just never going to happen anyway. And 37 is way too late to learn the keyboard or reaper or something from scratch. So I just have to live with this unsatisfaction.
I met you in secondary school computer classes and I found, I thought, a kindred spirit. Just like me you were physically uncoordinated, bookish, shy and filled with awe at the potential of these machines, the worlds they could create. Just like me, you were bullied by the popular, the sporty, the attractive. We talked of fantasy and science fiction and the small-mindedness of the adult world.
I met you again in university, many more of you. We were young adults now, and we nerded out about compilers and data models and virtual realities. We consoled each other about our absent love lives, discussed our futures, and enthused about how technology could transform and save the planet.
I appreciate all the earnest folks who are explaining why their institution's DEIA program was good, what impact it had, and why the govt eliminating it is awful. Really. I do.
But you're communicating the wrong thing. Because that's not the fight you're in.
The fight you're in is about segregation. The government is forcing you to choose segregation -- the obstruction of non white people (and white women) -- as well as queer, trans, and disabled people -- from entering or advancing in your industry.
You need to start clarifying THAT. Be. Clear.
Help the media and public call a Thing a THING.
Hold your industry accountable by CALLING THE THING THE THING.
Make that room, that meeting, effing uncomfortable.
And force the question: Are we choosing segregation so that we don't end up on a fucked up list monitored by some 19 yo incel?
For absolutely no discernable reason, NVDA has stopped reading the contents of new terminal windows until I alt+tab away and back again. I reinstalled Windows Terminal; no luck. I also have huge memory leaks again, which is something I haven't seen for a long time. I'll try TerminalPreview but am not holding my breath. I tried to switch back to conhost but I just can't. The output is read in an incredibly broken way. I genuinely have no idea what to do to get a good terminal experience anymore, or why it changed. There was no Windows or NVDA update coinciding with it. There are no NVDA errors when the terminal text isn't being read--it just ... isn't. The review cursor still works.
@jcsteh Oh, odd. I had it set to default, and I thought that was "notifications", but it's "diffing" now. I've specifically set it to notifications and according to my initial test at least, it's working. I wonder if (and why) the default state changed randomly. That could explain a lot.
@jcsteh The default never was notifications in mainline NVDA. Maybe I did that in one of my self’-signed integration builds? Until we have a good strategy for password suppression and handling of long output, diffing is my recommendation at this time.
@bill I've had similar problematic experiences with diffing, though, where text just doesn't read at all. That's actually why I tried Terminal and gave up in horror a while back. It took switching to notifications for me to realise Terminal wasn't awful. Something is definitely very wrong with diffing in Terminal.
@bill @jcsteh I know there's one particular nasty edge case when using diff match patch. Sometimes the process dies for some inexplicable reason. When this happens, any attempt to navigate, not just speak, anything in terminals generates an immediate error with some win32 error code about an invalid handle at the root of it, instead of simply trying to restart the now dead process when that exception occurs.
@x0 @jcsteh Yeah, all of this is really unfortunate. Because DMP is Apache licensed and NVDA is strictly GPL2, we can’t run them in the same process legally. Hence the nvda_dmp utility.
@bill @jcsteh Wasn't there also an issue of anything that was UTF-16 outside the BMP causing it to hard die? Can't say I've ever seen that so they might have fixed that one. Sort of wondering, wasn't the idea to switch from difflib to DMP based on performance? Does IPC not negate that? The solution to my particular issue is just catch the exception and if it looks like that, try just restarting the process? Yes, you lose some diff state, but that's better than *that*.
@bill @x0 Yeah, if there's one thing the Firefox Cache the World project taught me, it's that modern IPC performance varies widely depending on what you use and how you use it. Lots of little synchronous IPC round trips (query, answer, query, answer, etc.) are really slow. But less calls (even if they're a bit larger) can be imperceptible to users. Even think about NVDA Remote over a local area network: it's pretty snappy, and that's going across machines and with all kinds of networking overhead that IPC doesn't have.
Two documentation edits related to the following documentation pages:
CURLOPT_SSL_ENABLE_NPN
Issue:
The use of a full-stop breaks the redirection when rendered on a webpage
Action:
Removed the fu...
Minetestbildung ist jetzt auf Matrix Immer wieder haben datenschutzsensible Leute beklagt, dass wir mit einem Open-Source-Projekt discord nutzen. Deshalb haben wir zunächst als Alternative einen öffentlichen Matrix Space gegründet mit verschiedenen Unterkanälen. Auch Videochat ist dort möglich. Erste Anfragen haben uns dort schon erreicht . Beitreten könnt ihr über diesen Link: https://matri… Vollständiger Beitrag: blogs.rpi-virtuell.de/minetest… ##minetest #luanti #matrix
Hi, Ihr erwähnt die synod.im App mit dem Hinweis, dass sie keine Altersbeschränkung hätte. Soweit ich das verstanden habe, gilt die nur wenn auch synod.im als Homeserver genutzt wird. Euer Space ist doch aber, wenn ich das richtig sehe, auf matrix.org beheimatet...? Kann sein, dass ich etwas durcheinanderhaue, @csett86@ruhr.social weiss vielleicht Genaueres.
The Gemini protocol is a lightweight alternative to HTTP positioned as a spiritual successor to Gopher. One drawback is that without reuse of TCP connections, every navigation takes several seconds on geostationary satellite Internet.
There is again a pull-request submitted to the curl project to bring support for the Gemini protocol. It seems like a worthwhile effort that I support, even if it is also a lot of work involved and it might take some time before it reaches the state …
So glad I checked before filing an issue. #Iceshrimp was absolutely crawling whenever I tried to load a thread with more than about 50 replies. Turns out the database upgrade overwrote my postgresql.conf somehow, and the values were terrible. Checkpoints were taking five minutes. Aww. Poor wiffle database! Daddy made it all better, now.
Beetje hardop nadenken over Cloud Kootwijk, een "reddingsboot" voor als je zonder Microsoft/Amerikanen wil kunnen communiceren binnen een club die over is op MS365. Dat valt nog niet mee. Je kan wel webmail ofzo onafhankelijk aanbieden, maar ieder document wat je opslaat gaat toch naar de OneDrive, en dat was de bedoeling nou net niet. Ook komt MS365 totaal tot stilstand als er sancties zijn (je kan niet meer inloggen), en kan je geen losse website meer bereiken. Tricky. Ideeën?
klinkt alsof we een soort export/import tool moeten gaan bouwen. Maar als o365 gebocked wordt, of je account gebokkeerd , zou ik niet weten wat er te doen is, je data is dan echt gegijzeld...
Country music star Sheryl Crow said she will donate funds from selling her Tesla to NPR amid President Trump’s intensified calls to slash federal funding for public broadcasters — calling out…
Pro tip: save the QR code somewhere and whenever you want to re-setup simply scan it again . Also great to help other not so tech savvy users by simply giving them a QR code!
@SanskritFritz for example put the manually created davx5:// URL into duckduck.go search field and add qr! behind it. It will display the QR for you which you can save or send somewhere else
@radasbona that is not possible yet since the internals of DAVx5 are spread between the system and davx5 and a lot can go wrong when importing data back. we are currently refactoring a lot of stuff and hope to provide better backup possibilities in the (not so far) future!
Nadeel van een wat hogere boom worden: boel mail van verwarden en van mensen die heel boos op *mij* worden omdat de wereld niet is zoals zij willen. En ik begrijp dat vrouwelijke sprekers & minderheden nog 10 keer meer ellende over zich heen krijgen. Dus ik heb het nog makkelijk. Maar als ik niet 123 antwoord, weet dat het even kost om door de mail heen te waden om de goeie stukjes eruit te pikken...
Ich hab's geschafft, FB zu löschen. Das ist ein Spießrutenlauf und das Herunterladen meiner Infos hat erst nicht geklappt, hat sich irgendwie in der Nacht aufgehängt. Nun ja, die Kachel ist auch weg, ich komme nicht in Versuchung, da nochmal drauf zu drücken.
maybe an LLM could help to write migration code to a usable format faster than doing it by hand -> like automate the process and have a test suite to make sure it worked
NPR's Mary Louise Kelly speaks to Deborah Rutter, former head of the John F. Kennedy Center for the Performing Arts, in her first interview since the board installed President Trump as its new chair.
The #riffusion#AI says you can upload songs, give it a prompt, and it will make a remix for you. Being in capable of doing audio editing myself, that sounded exciting! So I grabbed an old @Onj track from his TBRN days, and asked for a hardcore dubstep remix. It gave me, uh, something. It's called Frankie Wolf, apparently. You remember that time you wrote Frankie Wolf, right Andre? Seriously, I bet nobody can even identify what track this was before the AI did unspeakable incomprehensible things to it.
@rlegowski1@Onj It deeply amuses me that the most identifiable thing it kept from the original track was the telephone ringing sound. Everything else came from, I don't know where. Some sort of AI fever dream.
I originally tried with Cheap Plastic Roger, but it just turned that into horrific shrieking noises with someone smashing a drum behind them. Completely unlistenable.
Its the day after valentines day. Also I was curious about how much effect my prompt has. I asked for the same track as a romantic love ballad. Why is it in german? This German man eventually gets just as excited about Frankie.
Right, one more and now I'm off to do something productive with my life. This time I wanted to be a bit more fair and ask Riffusion for something at least vaguely musically similar. I asked it to turn the track into a kpop song. It, uh, just kind of ignored me. Anyway, here's the Bollywood Freaky Freaky Fwoof remix everyone always wanted!
You should. It's free while you're in beta. It also advertises that as you favourite songs, and upload your own tracks, it can train itself to better match your musical style. So...I...ummm...I guess I'm pretty screwed.
Also, I'm not exactly being fair to it by giving it a track as strange as Freaky Freaky Fwoof to work off of. I suspect that people who aren't trying to be funny could make something pretty okay.
Did you maybe give it the instrumental version, not the version with the TTS voices and Patrick and the sheep and the ripping noises and the telephone and everything? Also, I don't pay for Suno. Does it let you describe what you want in a cover? For reference, this was the track I gave Riffusion.
Wen you pay, you can upload a track and have it covered in a variety of styles, give lyrics or specify instrumental, etc. When you use the free account you can only extend a track which is a different thing.
Oh, interesting. In riffusion you upload, then you just get a textbox to type your prompt. You can change the lyrics and pick instrumental if you want. But you don't have to select a particular style or anything. Just type whatever. You can also extend audio, but I think there's maybe a bar I have to drag or something? I haven't figured that part out yet. It would explain why Suno makes so much more sense if the input is more restricted.
Does suno try to transcribe the lyrics for you and do it really, really badly? These...are not the lyrics to Believer by Imagine Dragons... [Verse 1] First thing first, I'm the singer of the world It's down to my think I'm one percent on my services I was broken for my age, thinking my friends would make me so out I think my parents would build them to be my turn to show me high, high
[Chorus]hey hey hey i oh oh hey hey is You're the face of the future, of the blood of the rain. Woo -hoo! The blood of the rain. Woo -hoo!
[Verse 2] hey maybe now you maybe not believe me Hey, I like the butterfly, don't let me wait. My life is to the mind, just for today.
I didn't, until I went looking for Freaky Fwoof. And found that, and a thing you recorded about how you're really really bored, and pop rocks.ogg, and many other glorious things. No, I'm not a data hoarder! Shut up! LOL If they didn't want me to keep everything, they wouldn't sell me 20 tb drives on Amazon.
It's a case study about a piece of code from the game _The Witness_ (actually, from the art editor for that game) which he rewrote with dramatically better performance.
Kickstarter link: https://www.kickstarter.com/projects/annarettberg/meow-the-infinite-book-twoThis was a presentation I gave to the University of Twente in e...
My summary is that he addressed the original code's poor performance in three stages:
1. He came up with a new algorithm that didn't get progressively slower like the original, and also produced better output. 2. He found that even the new algorithm's inner loop was calling into code that did lots of unnecessary work, including memory allocation, so he instead wrote code that did something much simpler. 3. While he was doing the rewrite for #2, he went ahead and used SIMD.
He had some provocative things to say about the dangers of code reuse and "modern" C++ about an hour and a half into the talk. And I think we Rust programmers have to watch out for these things as well. It's easy to casually use high-level abstractions that aren't appropriate for the situation. I don't think we have to give up Rust's safety to avoid that problem, though; we just have to be aware of what the functions we're calling are doing.
Yesterday while watching one of those song identification YouTube videos, I encountered an artist going by the name “JVKE”, which is pronounced “Jake”. I noticed that VoiceOver does not pronounce his name “Jake”, it says “J V K E”, as one might expect. If I don’t capitalize the letters, I get a single syllable where the v is almost silent. It got me thinking about the times Apple has taken it upon themselves to “fix” the pronunciations of stylized names. A good example is the Canadian rapper “BBNO$”, which Apple correctly substituted for “Baby No money”. But what are the real-world consequences of changing the pronunciation of a stylized name, instead of leaving it alone? Let’s use my recent discovery of the JVKE song as an example. The stylized name is meant to be seen and not heard, so I can understand why changing the pronunciation might make sense. But I’m still consuming written words, just using a different medium. What if I then go on Reddit and talk about the “Jake” song I just found? No-one would have any idea what I was talking about. Some astute commenter would say “Wait, do you. Mean JVKE?” Which my phone would then read as “Wait, do you mean Jake?” If I knew to watch out for this sort of tomfoolery from Apple, I would probably check the spelling at that point and be suitably annoyed that my phone had tricked me into publicly misspelling the name. If I didn’t, I might reply to “Do you mean JVKE?” With “Yeah, isn’t that what I said?”. This is not a win for anyone. Since my phone did not change the pronunciation of the name, I got to have an experience that is probably quite familiar: I had to learn the pronunciation myself. If I cared enough, I would change it in my phone’s dictionary, but I’m not going to. I’m oversimplifying a bit here, but I feel like there are essentially two different camps of text-to-speech users—those who want things read to them and those who want to interpret. I want to interpret. I don’t want my phone to expand abbreviations or change names so they’re pronounced correctly, because then how would I know whether someone wrote “BTW” or “By the way”; “Washington” or “WA”; “Jake” or “JVKE”? I’m curious to hear other perspectives on this, if they exist. The main counter argument in my head is “reduced cognitive load”, which is completely valid. Would you rather your phone just “fix” all the common mispronunciations for you, or would you rather learn about them yourself and make the decision on whether to fix them, even if that means being potentially confused by stylized words that break the laws of the language?
English makes this a lot more complicated because there's no obvious place to draw that line. You can't just say "rules only, no special casing of words", because most users want special-casing for words like Louvre or Shadenfreude. It's just a question of where the line should be drawn.
Languages that only have a few simple, unambiguous rules for translating text to phonemes can get away with "rules only".
I think abbreviations break more than they fix, honestly. Orpheus is good about that, it doesn't bother with very many at all. Mr and Dr for example, don't get expanded. They are what they are. I mean is it Doctor Jake or Drive Jake? Or Jake the Doctor or Jake the Drive? Also St. Is it Fish Saint or fish Street?
I agree, but their should be an option to turn that dictionary on or off in accessibility settings. Also at least with Vocalizer, JVKE is pronounced as Jake when used in a sentence.
. Also great to help other not so tech savvy users by simply giving them a QR code!
Chris 🌱
Unknown parent • • •@doomsdayrs That's awesome :D
I really feel like having my own place will give me more freedom to socialize or do whatever on my own time