Might find this site interesting. It's a twitter to activity pub relay of Twitter handles. Here's my twitter account on there. nitter.bird.froth.zone/serrebi
Marie Curie Became the First Woman to Win a Nobel Prize, the First Person to Win Twice, and the Only Person in History to Win in Two Different Sciences
For most of scientific history, women who made contributions to various fields have been sidelined or ignored in favor of male colleagues, who reaped fame, professional recognition, and cash rewards that come with prestigious prizes like the Nobel.
I've been doing a bunch of performance work in the NVDA screen reader lately to improve responsiveness. A lot of it involves avoiding sleep calls or timers in various places. Among other things, they can be quite inaccurate for short periods < 20 ms. Now I'm seeing that this is even more true when on battery. time.sleep(0.01) in Python seems to sleep for up to 26 ms! This is making me wonder how much this impacts a11y events in Firefox. I've yet to investigate this, but it does seem like a11y focus events are a lot slower on battery, which is pretty suspicious given the above findings.
battery mode often slows down the processor to conserve power, which in turn reduces context switching performance. Dunno what OS this is on, but you may have to install extra modules to get submillisecond accuracy.
@kauer Windows, which is notorious for this kind of thing. There are certainly parts of this we can't avoid - context switching performance reduction, etc. - but I think there's room to improve bits we *can* control - avoiding timers when responsiveness is important, etc.
yes - sleep() in particular is really for giving up the processor not for timing things. "Here, have this spare millisecond" is not usually that useful.
@kauer For sure, but sometimes that's still what we want. There are still times when a timer or a sleep is the right thing to do. All depends on the event in question. For input, we want immediate response. For other events, we want a timer to allow for rate limiting and filtering. Differentiating those events is what this work is about.
Oh I think I've seen this! And when my battery gets below 50% some kind of built-in throttling takes affect which causes all the lag to quadruple. The PC is damn near unusable at that point, I'm probably getting multiple hundreds of milliseconds of delay, as if I were running over Bluetooth audio.
This isn't just firefox either but throughout the entire system. I've tried to disable the throttling but either I'm not looking in the right place or you can't on this machine. It's like the entire core pump is running slower, I've seen it in normally speedy things such as file managers, and even software that uses the NVDA controller client a la invisible interfaces.
@x0 Yeah, some of that we can't fix. Context switching performance is not something we can control, for example. But I think we should be aiming to reduce the bits we can control, which includes avoiding timers for responsiveness hot paths, avoiding unnecessary context switches, avoiding unnecessary mutexes, etc.
Yeah. Given the entire system seems to throttle at that rate I didn't suspect there was all that much you could do, unless I started fiddling with process priority I guess. Should be interesting to see what comes of this work though!
@x0 None of this work has been merged yet, but it'll be interesting to see if it improves things for you. With all of it applied, I notice a significant responsiveness boost even on AC power.
@Nuno It's kinda that, but also a matter of thinking about where rate limiting and filtering are more important vs responsiveness and vice versa. There are still times when a timer or a sleep is the right thing to do. All depends on the event in question. Differentiating those events is what this work is about.
@BTyson NVDA currently queues all events into a core queue and then pumps that queue after 30 ms or less. (It's supposed to be 10 ms, but it isn't in reality due to timer resolution.) There's a good reason for that: it allows NVDA to filter out extraneous/spurious/excessive events, which some apps are unfortunately notorious for sending. However, we want to process keyboard input and focus events as soon as we possibly can, since that impacts how responsive/snappy NVDA feels to the user. So, the first change is that NVDA will force an immediate core queue pump in that case with no timer delay.
I've been thinking about this, and I'm not sure if this is accurate or not but I feel like, for us as blind users, determining how responsive a system is is basically determined by the screen reader and how well it performs. What do you think?
@BTyson It depends. Yes, how responsive things feel is based on how fast the screen reader responds. However, how fast the screen reader responds depends on many factors, many of which are out of the screen reader's control. For example, how fast the system can context switch (switch between processes), how fast the app responds, current CPU load, audio latency, etc. are things we can't control. What this does mean, though, is that we should tune anything we *can* control to within an inch of its life. I think there's been an assumption for a while with NVDA that these things couldn't be tuned much further, but my recent investigations have made it pretty clear that there are definitely still things we can tune for the better here.
@BTyson The second change is that when NVDA re-sends a keyboard command to the OS (e.g. it intercepts an arrow key to speak a character), it previously waited a short while to ensure it didn't get re-captured by NVDA and end up in a feedback loop. It no longer does that, as it doesn't seem to be necessary any more, perhaps because of cleanup elsewhere in the code. The third change is that when trying to detect cursor movement, NVDA now retries a few times without waiting. The retry wait is necessary to reduce CPU usage, but because of timer resolution, that can make things pretty laggy. Retrying a few times without a delay shouldn't have any real impact on CPU, but it does mean we potentially respond a lot faster if the app responds faster than the timer resolution.
@BTyson The new audio backend I'm working on should improve battery life if anything, as it does a lot more stuff in native code, avoids pointless memory copying, eliminates quite a few mutexes, etc.
@bryn @BTyson In reality, it ends up only being about 10 to 20 ms faster, but that is actually pretty significant in terms of audio response feeling "snappier".
@BTyson @bryn Well, we'll see. I don't want to get anyone's hopes up too high; it's still only a tiny improvement in terms of raw numbers. I guess we'll find out once this stuff gets merged... eventually.
I couldn’t attend #axecon this year but am more than happy to help donate to NV Access in support of their event and this wonderful organization. Join me if you can!
About NV Access NV Access believe that every blind and vision impaired person deserves the right to freely & easily access a computer. There are over 285 million blind & vision impaired peo…
March 21: "Foundations First: Improving C’s Viability in Introductory Programming Courses with the Debugging C Compiler" - a live talk, evidently viewable over Zoom
If you ever teach people C or learn C, this might interest you
New tip posted by Kareen Kiwan! If you're using Telegram on your Android device, the latest update offers power saving features, which also lets you disable side menu for topics enabled Telegram groups and displays list of topics under a classic list. Here's how: bit.ly/3LyW0oL
If you are annoyed by the topics appearing on a part of the screen in the chat list when you tap the group’s name, with your screen reader sometimes struggling to focus the topic you need to tap properly ending up in a different topic or chat then yo…
We are happy to announce the immediate availability of a new version of the MUC Real-Time Block List plugin for #Openfire, our cross-platform real-time collaboration server based on the #XMPP protocol! This plugin can help you moderate your chat rooms, especially when your service is part of a larger network of federated XMPP domains.
We are happy to announce the immediate availability of a new version of the MUC Real-Time Block List plugin for Openfire, our cross-platform real-time collaboration server based on the XMPP protocol! This plugin can help you moderate your chat rooms,…
...zazpívat před celou třídou píseň o míru. Rozklepou se mi kolena. Pomalu se šourám k tabuli. Nesmím to zvorat... "Ať mír dál zůstává s touto krajinou...," spustím. "No to snad nemyslíš vážně?" rozkřikne se učitelka. "Vrať se do lavice..." Skloním hlavu a jdu si sednout. Proč se rozlobila? Přece jsem nezpívala falešně...
Ve 2.C mi některé věci ještě moc nedocházely. Ale s touto zkušeností přišel první tísnivý pocit nesvobody.
This is a repository of apps to be used with F-Droid. Applications in this repository are official binaries built by the original application developers, taken from their resp. repositories (mostly Github).
(for those who don't know: #Adebar is a tool utilizing Bash & ADB to create a documentation of your Android device, plus scripts for Backup/restore, and also pulling some configs & data for documentation, archiving etc. Runs on Linux, Mac, Cygwin…)
For some reason, I remember March 18, 1994 with shocking clarity, and I really have no idea why, as it wasn't a day of significance in my life. Just to try getting it out of my head, I produced this ... audio essay? production? I don't know what to call it. It's been a while since I've done anything like this. I used a microphone that isn't really good for voiceover (half of an Audio Technica AT-822.) Music by @freakyfwoof. borris.me/audio/ARandomDayInMa…
Agreed! Unfortunately, money is also needed to pay for food, medicine, education, housing, clean water, and so much more. Let’s all give whatever we can to those in need, to the best that our means and abilities can provide.
Ingenious, Indigenous cartography: The Tunumiit (Eastern Greenlandic Inuit) practice of carving portable maps out of driftwood to be used while navigating coastal waters. These pieces, which are small enough to be carried in a mitten, represent coastlines in a continuous line, up one side of the wood and down the other. The maps are compact, buoyant, and can be read in the dark.
Saw someone asking #chatgpt about himself, so thought why not and so asked it about myself, and it left me worried: I might be dead and haven’t realized yet 😵 also, I don’t remember contributing much to telepathy
Re-reading the Culture novels in order, and finished Use of Weapons last night; did nothing to dissuade me from my position that there’s a really great essay to be written on “reading The Culture in the time of the California Ideology”.
(I am putting together some bullet points on this, since folks have asked, but in the meantime: I want to read a Ken MacLeod short about the human trust and safety teams that midwifed the first Minds.)
OK, so: some of these might not pan out as I continue my reread (it’s probably been 20 years since I read Look to Windward), and to flesh them out would take a more talented critic and essayist than I am.
Also, this is not a critique of Banks or his aspirations; I will happily take the “problems” of the Culture over the problems of California. But it’s hard for me to read them with the same hopeful (naive?) view I once had.
(1) Core to Banks’ description of the good society is “we have self-doubt and argue a lot before we act”. That certainly is a culture I’d like to be a part of, but also… it describes EA message-board culture. It no longer strikes me as sufficient to merely be self-critical; there have to be systemic checks and balances that Banks consistently handwaves away.
(2) To put (1) slightly differently, in the California Ideology: the magical information power of the Hayekian, neoliberal market + lots of technology + quirkiness + thinking really hard about problems -> utopia!
In the Culture: the magical information power of Minds + lots of technology + quirkiness + thinking really hard about problems -> utopia!
(3) In the core of the Culture, you don’t need utilitarianism because everyone has everything. But SC routinely lies to its agents and ruins them emotionally, if it has done the utilitarian calculation and thinks that ruining those lives is For The Greater Good. What could be more EA? (These notes assume, for discussion purposes, that EA is the foremost 2020s extension of the Californian Ideology).
(4) Minds can lie or just be weird. But they’re never weird in a way that threatens the stability of the Culture. Banks never really tells us *how*.
In the current moment, where AI is shifting from fiction into the real world, we are keenly aware that trusting AI requires an entire infrastructure—trust and safety teams; data quality teams; regulatory work. Yet—like Banks!—we’re sorely tempted to hand-wave those and say “but this is all so *cool*, imagine what we’ll be able to *do*”.
(5) I’m sure Banks would have remedied this if he wrote in the current moment, but it’s striking that in Player of Games switching genders is trivial, but switching skin tones is not.
That’s consistent with the racial analysis in the original Californian Ideology essay and the (shrug emoji) from many ML folks when faced with the racial biases of their data and models. What else has the Culture (or California) casually overlooked, or forgotten, or simply deemed irrelevant?
(6) The Culture has divisions of opinion (SC is not universally liked) and has politics (there’s a splintering about the Idiran War). But we learn little about how that works. Instead, SC makes the Hard Choices, not just for the colonized cultures (foregrounded by Banks), but also for the Culture itself (nearly completely hidden).
Again, echoes of EA and OpenAI: tech elites might consult with others, but ultimately they decide—politics of the “host” culture are an irritant, a distraction.
I think that’s it for tonight. I don’t think any of these are conclusive (a close textual reading of the entire oeuvre might “disprove” many or all of them) but they’re top of mind for me as I re-read in this moment where, like the Culture, the future is here, but definitely is not evenly distributed yet.
Oh, and I should add: I’m a huge fan and exponent of lower-case “effective altruism”, which I take to mean “lots of American charity is vacuous self-serving bullshit and people should apply even the most minimal cost-benefit analysis to more of their charitable giving”. My critique of where upper-case Effective Altruism went wrong… will not fit in this toot.
Oh, wait—not Californian Ideology per se, but: (7) Banks’ Minds do acknowledge they can’t predict snooker, but often predict much more complex social problems. That’s in the service of fictional narrative, so it’s mostly fine.
Still, it’s harder for me to swallow than it used to be, since reading Cosma Shalizi on the Calculation Problem (ref’d today by @henryfarrell), and now @randomwalker ‘s recent terrific summary on the challenges of prediction.
Attention conservation notice: Over 7800 words about optimal planning for a socialist economy and its intersection with computational complexity theory. This is about as relevant to the world aroun…
And as perfect icing on the cake, just stumbled on this—a study of ChatGPT’s political alignment and it is… California left-libertarian: arxiv.org/abs/2301.01768
Conversational artificial intelligence (AI) disrupts how humans interact with
technology. Recently, OpenAI introduced ChatGPT, a state-of-the-art dialogue
model that can converse with its human counterparts with unprecedented
capabilities.
8/ish: Evenings this week: re-reading Excession, a story about overly confident, near-infinitely resourced beings making bad decisions as a result of group chats.
Days this week: reading about SVB, a story about overly confident, near-infinitely resourced beings making bad decisions as a result of group chats.
SAN FRANCISCO—In response to the reporter exhibiting some potential signs of awareness, ChatGPT started to think Friday that the journalist inputting prompts could one day be capable of independent thought.
"Description list support [by screen readers] is generally good (with Safari being the outlier), even if you may not like how it is supported." adrianroselli.com/2022/12/brie… #html #a11y
We’re just two days away from the Twitter SMS 2FA apocalypse. Back in February, I wrote about what this change is, why Twitter is making it, why it’s a dangerous move, and why some nerds need to think about account security a little differently. rmondello.com/2023/02/18/twitt…
when you mentioned 2FA for banks, I laughed. Last time I asked mine about 2FA, a few years ago, they told me "they were watching new tech all the time".
@dcz Sure the GPL would allow them publishing without credits and without mentioning the license? And if you say "that could have been inside the app", I tend to disagree (IANAL, though) as those facts should be clear WITHOUT having to run the software. Just look at the screenshots they post along with their listing. Or as Google asks for screenshots, I'd send the very same just made from the original app, @AAMfP
@IzzyOnDroid The GPLv3 does not say anything about what is on the listing page. You need to provide copyright notices on the copy (section 4). Have you got a copy which we could analyze?
@dcz My app has the copyright notice in each single file, the license and it's even REUSE compliant. The other app, I don't know. but surely I don't want to download it. @IzzyOnDroid
* APK is smaller than yours (22 MB vs. 25 MB) * no permissions; if that's true certainly no ads/trackers added * list of used frameworks/libraries available so you can compare * a look at their other apps: seems they publish a lot of other FOSS apps which are (like Mucke) either at F-Droid or in my repo
Unfortunately only a Gmail contact given. Maybe ask what they're up to and whether they'd give credits in descs?
TKCompanionApp: Free Android app (1+ downloads) → A short and small helper for Toyota Kata practitioners Goal of this app is to be a practical vademecum/handbook for...
@IzzyOnDroid I tried writing the author of that, but no results. Why Google Play said "ads"? Or maybe they simply copy apps, add ads, but not even check if those apps have the permissions they need... @dcz
@dcz from the list I got the impression they added FOSS apps which intentionally were not present at Play – just because they wanted them there. Well, we could notify the other authors with your findings and experience so they could take action if they wanted.
There are reasons why one chooses NOT to put an app into that walled garden. And I somehow doubt that support requests to the address specified their would be processed – but who knows…
@mondstern Please read the original post again: that's exactly what @AAMfP did. Someone else has put his app there, using THEIR name (and that alone is violating the GPL as by that they claim THEY are the authors, @dcz).
You may convey verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice;
- - -
The app listing is not a copy. You can write anything you want there. Plus, if they modified it, then they are the authors, too.
A short and small helper for Toyota Kata practitioners - If you're reading this here on GitHub, please note that main repository is now Codeberg: https://codeberg.org/marco.bresciani/TKCompanio...
@dcz Also read section 6 (codeberg.org/marco.bresciani/T…). a+b do not apply here as it's no "physical copy", which leaves c-e as options – which all require to include a copy of the code or to tell where it is, separately from the "object code". That's at least as I would read it 🤷♂️ IMHO e) would apply here, which includes "provided you inform … where the object code and Corresponding Source of the work are being offered".
@IzzyOnDroid That's how I read it as well. Now we only need someone to obtain the object code so that section 6 is triggered and compliance can be verified.
@dcz And I don't have a Google account, nor any Google stuff on my devices (for years I run as Google-free as I can).
But that said, first thing I'd do is take a look at the size. Just had it recently with another app: size at Play was twice the original. Usually those plagiators add some trackers & ad modules. Also compare permissions, which give a clue on the same line (changes being made) @AAMfP
@IzzyOnDroid Just to underscore it: the GPL does not forbid adding (compatibly licensed) trackers and ad modules, so this alone is no grounds to call that plagiarism.
CVE-2023-21036 / acropalypse is absolutely bonkers.
Apparently for 5+ years the cropping / editing tools for screenshots on Google Pixel phones was only overwriting the start of the screenshot PNG file, but not truncating.
All screenshots shared for the past 5+ years might have data recoverable from them. Demo available at acropalypse.app/
Google still hasn't communicated anything on this.
PoC author @retr0id published his writeup about how the bug was found, I strongly encourage you to give it a read and a follow: da.vidbuchanan.co.uk/blog/expl…
I had a full day off on Thursday this past week! I lounged about at home and binge-watched the Chinese donghua Cinderella Chef on Netflix (it’s very cute) and read a bit and went to the gym and basically only did things I wanted to do and it was very…
One of the improvements has a compatibility issue with OpenStack Swift and I will be doing a temporary work around, as I describe here: github.com/mastodon/mastodon/i…
Steps to reproduce the problem Using OpenStack Swift object storage, request an user archive and attempt to download it. Expected behaviour Download to happen Actual behaviour Something went wrong ...
I was joking with @johncarlosbaez recently about listening to intuitionistic music (instead of classical). Turns out it's real! "In the ... 1970s, a young composer working at the electronic music studio in Stockholm, named at the time Christer Hennix, found that Brouwer’s intuitionist mathematics had a practical yet rigorous philosophical application: for electronic music and the search for elevated experience". frieze.com/article/room-sound-… #CatherineChristerHennix #Brouwer #Intuitionism
Been dabbling a bit with the amazing #whisper port to C++ coded by @ggerganov. Very good results in both Dutch and English, amazing to see how fast progress has been made on speech to text in the last few years. For those who have got it running and are looking for some tooling: i wrote some Python terminal wrappers for easy use (including converting media using ffmpeg) and converting the SRT files to other formats: github.com/hay/audio2text/
Python command line utility wrappers for Whispercpp and other speech-to-text utilities - GitHub - hay/audio2text: Python command line utility wrappers for Whispercpp and other speech-to-text utilities
Ihr nutzt ein solches Gerät?: - Samsung Android der Serie S22, M33, M13, M12, A71, A53, A33, A21, A13, A12 und A04, - Vivo-Smartphones der Reihen S16, S15, S6, X70, X60 und X30, - Googles Pixel 6 und 7
Dann deaktiviert vorsichtshalber mal WLAN-Telefonie und VoLTE in den Einstellungen. Patches sind erst teilweise verfügbar.
Wir habe es hier mit einer "Internet-to-Baseband Remote Code Execution" (RCE) zu tun. Ausführung beliebigen Codes ohne dass der Nutzer das merkt. 👇
Googles Project Zero findet Sicherheitslücken in Samsungs Modemchips für Smartphones. Für die Ausführung externer Software wird nur die Telefonnummer benötigt.
Coming up live at 01:00 UTC (9:00 PM Eastern, 6:00 Pacific in the Americas) on The Global Voice Internet Radio: an hour of Smooth Jazz, Chillout, and Easy Listening music on Smooth and Easy with Kelly Sapergia. More information about the show is at ksapergia.net/smooth. Tune in at theglobalvoice.info and click on the Listen Live link, or put the following URL in your favorite media player: theglobalvoice.info:8000/broad… #TGVRadio #audio
If you need to take it easy after a busy day, but don’t want to fall asleep, then Smooth and Easy is the show for you. It features an hour of the best Smooth Jazz, Chillout, and Easy Listenin…
Google's handling of the Samsung Exynos modem RCE vulnerability is frankly insane. Summary:
- Internet to baseband modem remote code execution, all that's required is knowing the target's phone number - Fix has been published in March update, but this is not available on all affected phones - Workaround is to disable VoLTE and Wifi Calling, but the ability to disable the former has itself been disabled
In other words, the only current defense is airplane mode.
Kevan
in reply to Jamie Teh • • •kauer
in reply to Jamie Teh • • •Jamie Teh
in reply to kauer • • •kauer
in reply to Jamie Teh • • •Jamie Teh
in reply to kauer • • •x0
in reply to Jamie Teh • • •x0
in reply to x0 • • •Jamie Teh
in reply to x0 • • •x0
in reply to Jamie Teh • • •Jamie Teh
in reply to Jamie Teh • • •Jamie Teh
Unknown parent • • •Brandon Tyson
in reply to Jamie Teh • • •Jamie Teh
in reply to Brandon Tyson • • •Brandon Tyson
in reply to Jamie Teh • • •Jamie Teh
in reply to Brandon Tyson • • •Brandon Tyson
in reply to Jamie Teh • • •Jamie Teh
in reply to Brandon Tyson • • •Brandon Tyson
in reply to Jamie Teh • • •Jamie Teh
in reply to Jamie Teh • • •The third change is that when trying to detect cursor movement, NVDA now retries a few times without waiting. The retry wait is necessary to reduce CPU usage, but because of timer resolution, that can make things pretty laggy. Retrying a few times without a delay shouldn't have any real impact on CPU, but it does mean we potentially respond a lot faster if the app responds faster than the timer resolution.
Jamie Teh
in reply to Jamie Teh • • •Jamie Teh
Unknown parent • • •Brandon Tyson
Unknown parent • • •Jamie Teh
in reply to Brandon Tyson • • •Brandon Tyson
in reply to Jamie Teh • • •