JPEG XL

Happy New Year!

https://www.instagram.com/reel/CmrCi99gvuv/?igshid=OGQ2MjdiOTE=

<:thinkies:854271204411572236>

I’m currently using a computer with 4 GB of RAM. I have a laptop iwth only 2 GB. They never really gave me trouble, they work well for anything I have wanted to use them for.

are there ffmpeg specialists here? I would like to ask a very difficult question about the video

Do you use 32bit os on that 2gb Ram laptop?

I do not. It seems prebuilt binaries are being decreasingly frequently provided for plain x86 nowadays.

at which growth rate? 😈

https://twitter.com/carlzha/status/1556095722738810880 Taiwan pizza...

No, I don't want eat 🐔ball 😢

https://github.com/libjxl/libjxl/pull/2023 would have been nice to match this with the New Year 2023-01-01, near success 😛

<@794205442175402004> since e10 is a thing now... <:KekDog:892418529117233162> ~~don't know if saving 6KB was worth it, but oh well~~

Hehe

probably will be slower for decoding need also bruteforcing only modes which are fast (or at least not slow) for decoding <:Stonks:806137886726553651>

мusic

https://www.instagram.com/p/Cmcp4lqPuJy/?utm_source=ig_web_copy_link

мassiv

https://www.youtube.com/watch?v=FCYFJ6BLApU

My boyfriend

Lol smokers <:KekDog:805390049033191445>

???

Left 1 is the original color, right one became brown, because people smoked inside

Oh

That's disgusting. Imagine how much darker it must get inside their lungs, where most of the smoke is directed.

I prefer watching more uplifting things on youtube, like this: https://youtu.be/qx8hrhBZJ98

https://youtu.be/Juqke35ZCLo

send link 😛

Boarding the plane to Dubai now, a short 6.5h flight. Then the long 13.5h flight to Sydney...

cool

Just arrived at the hotel in Sydney. They accidentally booked me for February instead of January 🤦 so it will take some time to rectify that...

when you squint a bit, they are same months 😉

Just a one-off

Anyway, it was corrected. Now time to sleep for me, it's already almost 2am here in Sydney...

Jetlag recovery day, the meetings start tomorrow...

here's the same image for those who are also in australia and want to view the image properly

https://imagemagick.org/Usage/resize/#resize_colorspace

Is resizing through the RGB colorspace instead of naive sRGB really worth it?

In general, what's generally the "highest quality" way to resize an image?

Resizing in linear is worth it, yes. Which downscaling algorithm is best, that's a tricky question. I am still investigating it. It's very much a "pick your poison" thing: you can make trade-offs between risk of blurring vs sharpness with risk of moiré/aliasing. It's a bit of an artistic choice, really. But doing it in linear space is definitely more 'correct'.

In this page, they mention RGB, LAB, and LUV. They go on to tout the benefits of each colorspace, then end up saying "it depends" as all of them are better for different scenaros. LAB and LUV even give similar results to sRGB at times. To quote from the page: > Nicolas Robidoux summed it up nicely with... > Generally, the linear light colour spaces (linear RGB and XYZ) produce exaggerated dark halos, and the "perceptual" colour spaces (sRGB, LAB, LUV) produce exaggerated light halos. > > If you think about it for a minute, this makes complete sense, because the perceptual colour spaces pack lots of bits at the darker end of the intensity spectrum, and "hollow out" the lighter end, so as to mimick the HVS (Human Visual System). So, 1 unit of dark overshoot gets you less "far" in sRGB than in linear RGB, but 1 unit of light overshoot gets you less "far" in linear RGB than sRGB. > > Sigmoidization (see next) treats dark and light overshoots equally, and generally dampens the extremes of both.

I'm getting a bit confused as to which is the best general method. RGB? LAB? LUV? Just sRGB? Is there really none, and it just depends on the image?

That explanation does not make a lot of sense with float computations 😛

Downscaling and upscaling are quite different. For downscaling, linear is best. For upscaling, I don't know, that's something that generally should be avoided anyway.

Float or int doesn't make that much of a difference, the main thing is how you average two colors: in linear space, the average of white and black (1 and 0) is 0.5 gray, while in sRGB it is something like 0.25 gray. Basically downscaling in sRGB (or any other gamma-compressed space) results in pixels that are darker than they should be, and potentially colors that are different from what they would look like when looking at the high-res image from a distance.

https://www.youtube.com/watch?v=8D-ZaO5pmHE

Is there a way to get libvips to do linear RGB resizing in 16-bit color space? It seems like it's doing it in 8-bit color space which is causing bad results... Sample commands: imagemagick naive (sRGB): `magick dark.png -resize x720 -filter lanczos magick-srgb.png` imagemagick RGB: `magick dark.png -colorspace RGB -resize x720 -filter lanczos -colorspace sRGB magick-rgb.png` vips naive: `vips thumbnail dark.png vips-srgb.png 1280` vips RGB: `vips thumbnail dark.png vips-linear.png 1280 --linear`

Original file:

magick RGB:

vips linear:

The blacks get crushed because of the 8-bit calculations.

<@310374889540550660> probably knows...

Is there a way to get what codecs the browser supports easily in JavaScript?

for images, iirc create an image using a blob

Would that list them, or would I need to then test them individually?

you need to test them individually

for example ``` const avifData = "data:image/avif;base64,AAAAIGZ0eXBhdmlmAAAAAGF2aWZtaWYxbWlhZk1BMUIAAADybWV0YQAAAAAAAAAoaGRscgAAAAAAAAAAcGljdAAAAAAAAAAAAAAAAGxpYmF2aWYAAAAADnBpdG0AAAAAAAEAAAAeaWxvYwAAAABEAAABAAEAAAABAAABGgAAAB0AAAAoaWluZgAAAAAAAQAAABppbmZlAgAAAAABAABhdjAxQ29sb3IAAAAAamlwcnAAAABLaXBjbwAAABRpc3BlAAAAAAAAAAIAAAACAAAAEHBpeGkAAAAAAwgICAAAAAxhdjFDgQ0MAAAAABNjb2xybmNseAACAAIAAYAAAAAXaXBtYQAAAAAAAAABAAEEAQKDBAAAACVtZGF0EgAKCBgANogQEAwgMg8f8D///8WfhwB8+ErK42A=" await fetch(avifData).then(r => r.blob()).then(blob => createImageBitmap(blob)).then(() => true, () => false) ```

Hi, vipsthumbnail actually uses float for linear resize

it's importing to scRGB using the image profile, so any crushing is probably a bad ICC profile on the image

(at a guess)

Your "dark.png" original has no profile, so I guess discord has stripped it ... maybe upload the original somewhere, or better still, open an issue https://github.com/libvips/libvips/issues (github is better for debugging than discord chat)

Hi, thanks for replying! I just took a look at the source code right now, so nice coincidence, I also just learned about the scRGB conversion. If I download the original dark.png file and run `magick identify -verbose` on it, I still get an ICC profile: ``` icc:copyright: Copyright 2019 Google LLC, CC-BY-SA 3.0 Unported license(https://creativecommons.org/licenses/by-sa/3.0/legalcode) icc:description: RGB_D65_SRG_Per_g0.50994 ``` Although, I'm not sure why it would need an ICC profile...

I'll try to get another sample image to isolate the "issue" I'm seeing.

It's actually a PNG that's converted from a JXL that was extracted by ffmpeg, so maybe that's causing some headaches.

libvips has two ways to make scRGB for resize: if present, it'll use the embedded ICC profile (the image might have some weird non-standard gamma, so this is the rigtht way to get linear data). If there's no profile, it'll assume sRGB gamma and fall back it its own sRGB->scRGB conversion The built in sRGB->scRGB conversion doesn't move the black, so if you are seeing crushing, I would guess it's probably a bad image profile

I tried your image like this: ``` $ vipsheader -a dark.png dark.png: 1520x855 uchar, 3 bands, srgb, pngload width: 1520 height: 855 bands: 3 format: uchar coding: none interpretation: srgb xoffset: 0 yoffset: 0 xres: 2.834 yres: 2.834 filename: dark.png vips-loader: pngload ``` So if there is an ICC profile, libvips can't see it. Again I'd guess discord has stripped it (but just a guess)

I'm not sure where the 1520x855 resolution you're seeing is coming from, the original file was 3840x2160. Maybe some weirdness with Discord. Also, it seems that `djxl` adds an RGB ICC profile when converting to PNG, so there's where that profile came from. Again, I'll try to be more thorough with this and then file an issue if it really is something with libvips.

Hmm identify can't see it either `identify -verbose dark.png | grep icc` maybe I'm using discord wrong? I clicked on the original, then right-click and "open image in new tab", then save ... is there a better way to download an image?

I think there's an "Open original" button somewhere that actually downloads the original file

Or "Open in Browser"

It's in small text near the bottom of the image preview, annoyingly

haha I didn't even see that, I'll try again

I think you're right, it might have been some ICC profile weirdness.

I tried again, I agree it's marking the background a bit darker, but it's not crushing things, as far as I can see

If I extract the frame to a PNG directly with ffmpeg, it doesn't have an ICC profile, and the magick RGB way of resizing things matches the vips thumbnail.

any idea where the profile came from exactly? is it supposed to be srgb?

It's what you get when you convert anything to JXL via ffmpeg, then convert that to PNG via djxl.

that's... weird

I'm getting such weird results converting things back and from from PNG to JXL via ffmpeg... and the extracted frame from ffmpeg is even different than the mpv screenshot

honestly im just lost at this point

maybe after this pr it will be better https://github.com/mpv-player/mpv/pull/10998

okay, that's weird, my jxl image sequence seems to have correct colors displayed (but still messes up vips ICC), but i can't replicate the issue extracting a single frame via ffmpeg

ill just debug it later when its not so late here

`g0.50994` or equals inverted gamma of `1.96101...`, that's an unusual value if its sRGB 🤔

I wrote something but I don't know if it's any good. Feel free to tell me what you think — did I write something obvious, cheesy, woke and condescending? Or is this some interesting and insightful food for thought? Your call: https://docs.google.com/document/d/1Il_eQOxAfPYt5Ru5aBVk9laTz3Hmtl6yWh3lWCHlnA8/edit?usp=sharing

(I genuinely don't know if this is an interesting text or not; if it is, then maybe I should put it somewhere online but if it isn't, I better just leave it at this)

(also: this is really very off-topic)

If you think it's crap, please don't be polite by saying nothing. I do standardization, I have a very thick skin, I can take your criticism and even appreciate it 🙂

<@456226577798135808> it's just two pages 🙂 But TL;DR: I'm arguing that the globalized human civilization should be seen as a single macro-organism, and that therefore we've collectively become "the last human on earth", and we better start watching our health.

I actually have a similar opinion, back when I used a concept of societal suicide years ago

if you see a society as a macro-organism, you can put every allegory that usually applies to humans to societies, and in some sense we are heading towards a voluntary extinction

in my country---South Korea---the birth rate is unbelievably low, _much_ lower than everywhere else, and reasons are usually described as a general (financial or emotional) difficulty in young couples raising kids and a broader reach of feminism values (for good or bad, because this also correlates to the rise of right-wing or trans-exclusionary feminism in SK)

https://tenor.com/view/games-vintage-pixels-gif-9131396

but none of them seem to be highly specific to South Korea! in some sense, SK may be experiencing some sort of inevitable fate much earlier than other modern societies.

is it natural or should be "cured"? I dunno, I'm a mere component of this macro-organism 😉

We're all just tiny cells of a big macro-organism, but I think we can strive to behave less like yeast cells and more like brain cells...

This game is pure nostalgia <:KekDog:805390049033191445>

I should port this to some less niche platform: http://repo.openpandora.org/?page=detail&app=nanolemmings.wb&p=1

Hm, does e10 work with lossless jpeg reconstruction?

idea: lemmings, but individually programmable 😉

no idea, but if it does, it would only affect the DC encoding and it would be wasting a lot of time recompressing the AC in exactly the same way hundreds of times

Ah, it does crash every time, but hoped it would reduce the size at least a little bit

Or well, it fails

maybe open an issue? it could be made to work, and might save a few bytes at a ridiculous cost

Well, I'm an aomenc 1 thread cpu0 user, so guess I'm used to ridiculous encoding times :p

There's an issue about it already actually

I was thinking to just disable E10 with jpeg recompression but I never managed to find the willpower to get into it 😛

<:SadCat:805389277247701002>

I mean, it's the same as e10 vardct: ideally the AC is encoded only once and the iterations are only done for the DC groups...

I really liked it! I think it was fairly encouraging (or incentivizing) of change, specially towards the end. It’s an interesting way to connect the two thoughts of “people in the world come together to form a single entity that works together” and “we should be self‐conscious of our future to try to avoid perishing”, which to me is a new take on the former idea that I hadn’t seen before. The analogies worked really well with, specially the yeast collony one, I feel — that was fairly thought‐provoking. I just felt like the introduction was a bit too extensive. I understand that the purpose was to simultaneously set the mood for the rest of the writing, as well as to bring back our understanding of science and biology that the text would then be using to analogize. I also feel like it was a way to succinctly show understanding of the subject, and thus competence on writing about it. Still, though, I feel like it could have been a bit less detailed. To me, it felt like it was just reiterating things I already knew from middle school, and I kinda skimmed through it a bit. Which I suppose makes it fine — if it seems interesting to someone, they can read through it, otherwise they can just skim through it like I did.

Thanks for the review/feedback!

I think I'll shorten the second paragraph a bit, it's only really there to prepare the reader for the later analogy that livestock/crops are the mitochondria of the macro-human, but it indeed doesn't need to be that detailed.

i disagree with the notion that there is only single macro-human left. evolutionary forces encoded deep in our genome naturally push for splits whether anything gets too monotonic, prime example being you having enough individuality to write up this article, and it's also very clear to see with ongoing wars and conflicts, even if you can't view them as a positive thing. and it's exactly the thing what will prevent total collapse, even though i can agree that right now things aren't heading in too good directions, but I'd say it's better right now than environmental harms what have happened 10s of years ago

sometimes it has to get worse first, before it gets better

I don't think total collapse in the sense that individual humans stop existing is very likely. But collapse of civilization as we know it (i.e. the death of the macro-human, even if some of its cells or organs might survive and eventually regenerate new macro-humans of substantial size and sophistication), I do think that's very much a possibility. History is full of examples of advanced civilizations completely collapsing.

Jared Diamond wrote an interesting book about that, 'Collapse'.

some amount of collapse and de-globalization already happened during covid tbh

and yes i'd say we're a bit worse off right now because of that

Surely there's more protectionism going on in trade, for example, but I think all in all, culturally and economically, on the long term, the globalization trend is unstoppable even if it has some ups and downs in the short term. I don't see us going back to almost completely isolated pockets of civilization like it was in the dark ages...

even in "dark ages" trade between countries existed

(that is, until civilization as we know it does collapse, which may be what we are starting to witness with these crazy past few years, but I think it's too early to draw the conclusion that this is actually happening right now)

individual parts of civilizations came and gone, but there's always been some contact

what we've observed during covid is regression

I mean in Europe, compared to the roman empire, civilization was significantly less advanced in the dark ages. A lot of science got lost and had to be rediscovered, the levels of economical organization dropped significantly, etc.

but like even in countries which have big cities, there are small, much less connected cities still, i think that people living in these don't feel such big globalization point of view, and only we living in bigger ones do

regarding knowledge, i don't think we're going back to book burning

anytime soon, if at all

Yeah, not anytime soon. And yes, in rural areas there are certainly 'backwards' people who are culturally/economically less strongly 'integrated' in the macro-human, so to speak.

It seems most likely that when civilization collapse happens, they are most likely to survive it the best.

indeed. and there seem to be forces for such structures to exist, ones who want to live in big cities just move there, ones who don't actually sometimes move out to smaller ones. smaller cities don't seem to move into directions of growing too connected on their own

i think humanity will be mostly fine unless we so severely mess up environment to the point where it gets harder to live for anyone, not only big cities but also rural ones and animals too

not seeing that happening in my lifetime though so it'll probably be fine, thinking too hard about that should be left for future generations

can't see too far into the future with limited information anyway

Humanity is probably almost as hard to kill as cockroaches or ants, but if civilization collapses into barbarism, it will be a severe regression and it could easily take us a millenium to get back to the level where we were. So I would still prefer it if we could avoid that 🙂

https://palant.info/2023/01/09/touchen-nxkey-the-keylogging-anti-keylogger-solution/ a glimpse into the nationwide security theater in SK

a side effect of moving from activex to "executables" some years ago, it is

applications only providing .deb and .rpm files is pretty common tbh

since you can't reasonably provide packages for everything

I dunno, I never used them in macOS or Linux, and in fact I refrained from using them even in Windows (as many things can be now done in Android anyway)

so in some sense we have moved from ActiveX to Android, and executable-based approaches are very much late (not as dead, but won't get you day by day)

no, but they now come with things like V3 Mobile Security which is almost as pointless in sandboxed environments like Android (I *am* aware that Android is far from being secure, but I doubt those products do help mitigating them)

so I'd guess those apps won't work well in non-typical Android builds

a preloaded security app? huh

anyway, at this point the infamous ActiveX-based certificate system is almost gone, but the root problem posed by it didn't really change

there are two major problems: it prevents usage from those are not "typical" Koreans, and it doesn't improve (if not degrade) security at all

yeah that's a typical thing in SK too, in fact you can't do anything without a mobile phone

I should've said a mobile phone *attached to Korean phone numbers* to be exact

cause companies now use phone numbers to do the verification

indeed

well to my knowledge in PRC a very small number of super-apps dominate the market, in SK not that much though

Kakaotalk is a super-app but not effective at keeping users in its shoulders, I was pretty fine without using Kakaotalk

Spaceship Size Comparison Chart 2023 https://www.deviantart.com/moreorlesser/art/Spaceship-Size-Comparison-Updated-V22-903959442

lol wtf? Why would you need to install a "security program" on your PC to log into your bank's website? Hell no.

That sounds as sketchy as calling the "Windows Security" hotline and giving them your credit card number at 1-800-LOL SCAM

I believe pretty every SK people will agree to your assessement

Does the South Korean government mandate banks to install spyware on their customers' PCs?

no, it's (supposedly) means to the end

That's hilarious. I can't believe this is real.

sk government already has spyware in everyone

I thought South Korea was supposed to be the one where people enjoy relative freedom compared to the North?

unfortunately the "security" industry has chosen a particular approach that is: i) actually plausible for a time (that ultimately originates from USA's crypto export ban, so SK had to roll their own 128-bit encryption scheme) and ii) easy to do by themselves but iii) ignoring virtually every customer comfort.

south korea doesnt have freedom

so when the original reason from i) was gone, they just kept everything as is until it was really, really impossible to keep

which freedom? ;D

should I talk about a recent development in SK politics... (nope, I won't)

isnt their thing where every online service tied to government id a government mandated thing

even that was not a mandate

Name a "country" or a "government" that does allow "their people" to have freedom.

true

:)

side note: if you stretch a definition of "government id", then you will have to have some "government id" for doing many things

a problem in SK was that, too many companies relied on the _actual_ government id

Show me your papers, citizen

Papers please

The royal paper

if you are in USA, it's like every company asking for SSN

Ah yes, tell me your number too

without any security consideration, an once popular BBS software had required SSNs by default

it was really crazy

Maybe we should tattoo your number on your arm or forehead, that would make it more convenient to keep track of you

eventually the SK equivalent of SSN was exposed... multiple times per every resident

Gotta make sure we responsibly keep track of everyone, for safety of course :)

It's my job to keep everyone safe

nowadays that much hilarity is gone, but you will be required to have a domestic mobile phone number instead 😉

which acts as an effective proxy to the government id, so to speak

But it's never my fault if something unsafe happens. It's my opponent's fault.

it always bothered me because something as not important as signing up for a video game from SK requires that verification

oh, that's another story

because an age verification is a thing in SK

I should note that I'm pretty much okay with an age verification and it is undesirable to show _everything_ to minors, even though I also agree that this boundary is and will continue to be very fuzzy

(the main reason I'm okay with that is that, you will eventually get to a threshold regardless of what you do, so it is meant to bother but not completely shield minors, and it works within its limitation)

anyway, since minors can always lie about their age, you are often required to verify your age in more sure way... which turned out to coincide with that thing used to verify yourself in websites

once an resident registration number (the SSN equivalent), nowadays a mobile phone number

not sure foreigners can even pass this process at all

Honestly everything would be a lot better and simpler if it was illegal for kids to be on the internet at all :P

The phone number bullshit is not for protecting kids since kids have phones too.

It's just to invade privacy.

of course you won't get a free pass just because you have a phone. it will be used to actually check if the phone owner is a minor or not.

Why would it be in their name? It's the parent that's paying the phone bill.

it's just privacy invading bullshit

parents will pay the bill but they will be tied to kids 😉

so, yeah, something is very off

to be fair, an alternative way to verify yourself itself is fine, and will be probably required for an effective age verification anyway. but the phone number and owner verification via telcos is kinda flappy at best

Honestly I would be totally happy if there was a new draconian law against kids being online in any capacity at all

back in time when the RRN was required, every kid knew their parents' RRNs for the obvious reason

now they will try to steal parents' mobiles lol

Parents don't parent

(don't know they actually do that, I haven't been a minor for decades...)

Anyways pretty sure USA does not have a monopoly on encryption libraries

There are lots of open source crypto libs that have no US export controls

not much in 90s

like the entire openbsd operating system

IIRC at that time not only SK but many other countries had their own schemes, only SK's survived because of a weird coincidence though

The 90s was 30 years ago.

30 years is plenty of time to install openbsd

not just openbsd but US crypto export ban was (practically) lifted many years ago

so yeah, I can agree to the original reason but I have a right to criticize everything else following that

No, it is still in effect

"practically"

and the modern cryptography doesn't involve much from NIST anyway

It is illegal for me to "distribute" crypto to countries on some list somewhere

That is still a crime for me to do

ah the axes of evil, so to speak?

but as I've said the modern cryptography allows you to completely bypass the ban anyway

I don't know, just ask the fedora project, they seem to be the most anal-retentively cautious about "following" weird US law

like software patents. I heard software patents are only a thing in USA and maybe Japan?

I wonder how the hell the USA manages to compete with countries that don't choose to willfully shoot themselves in the foot with the stupid concept of patents

more like that those are jurisdictions _known to have_ a court-verified notion of software patents in some way

I think you can make software patents a thing in many enough countries if you really wish

(and you have a source of infinite money and lawyers)

Patents are an inherently stupid idea. Software patents are taking a stupid idea to the most stupid logical extreme.

at least patents expire after 20ish years

let me talk about copyright extension...

Copyright is another inherently stupid idea. It's basically a less-extreme extension of what patents are, and patents are a far more extreme and more generalized idea.

both patents and copyright started out as a practical way to protect rights, and then they got ridiculous fast

they should be shortened quite a lot

especially patents completely ignore the speed of the modern society

20ish years is an entire generation, much longer than you can actually make your idea a reality with a reasonable funding

if you can't make it a reality in, say, 5 years, do you really think you can make it true in following years

the same can be said to copyright, how on earth why does it depend on the death of the author

that alone effectively enlengthened the copyright by 30 years

I can understand the idea of "don't lie about taking credit for something you didn't come up with" but copyright and patents are "I believe I own something intangible and therefore that gives me the right to control the use and spread of information itself. I am allowed to control other people at a very personal level to ensure they only use the information I gave them in approved ways."

And that's just delusional insanity

All of these ideas are.

side note: copyright in US used to require a registration, I don't recall when it became automatic (late 20C?)

so much like the patent office (in the past, now it really doesn't) there was a chance to sort out the "actual" ownership

in the other words they do not simply scale and force-scaling them will result in a very fragile system

which we have today

Before they were established into law, it was recognized as privileges, not a "right" but a privilege the government grants as a reward designed to incentivize and protect artists or inventors

yeah

Nobody has some natural inherent "right" to protect information the same way as protecting a tangible thing like you protect your home or your person

But some people forget that and think it's some "natural right"

I don't have anything to say for or against the idea of "natural rights" if you want to argue that you have a natural right to defend your person or your home from an armed assailant...

But if you try to extend that to intangible ideas and information, that's delusional.

they seem to think it as some sort of property right? (you know, the "information property" reflects this reasoning) and of course they don't realize it is artificial, that can be bent and tweaked as the society agrees to restrict those "rights".

side note: that can be also said to real estates, but that's too much for this server 😉

I just want to be free, like everyone else does.

But most people in this world want to serve their Lord, and/or have everyone else bow down to them.

Maybe a tenth of a percent of people, or maybe even less than that, believe freedom is better.

After all, freedom is out of control! It's not safe!

https://www.youtube.com/watch?v=Iqo4bXq1dM0

https://xkcd.com/2726/ lol

it's recursive

For checking if a browser supports encoding an image, there is the `canvas.toDataURl()` function, which will try to encode in the given MIME type, but return png if not supported, but that runs the risk of browsers that support decoding but not encoding (or the opposite for some reason) not being detected.

It's just annoying that the video codecs have https://developer.mozilla.org/en-US/docs/Web/API/HTMLMediaElement/canPlayType, which allows you to just ask the browser

https://blog.google/inside-google/message-ceo/january-update/ this one really makes me go "WTF, Google?!" even more than the jxl-in-Chrome decision.

Not making enought money? <:Thonk:805904896879493180>

https://news.stanford.edu/2022/12/05/explains-recent-tech-layoffs-worried/

Amazon and Microsoft also cut 10k+ jobs already in 2023. It's insane. For no good reason, just because the others are doing it so now the big shareholders think it's something they need to do.

I do wonder how much of this big corporate behavour is the same small group of executives/investors in their own bubble devoid from reality.

my company also had big layoffs this week from cancelling a main product which everyone was mad about

they also removed almost all the contractors

Even if you want to refocus as a company, I don't think layoffs are ever a good approach to that. People can refocus too and work on different things. It's just irresponsible to do refocusing via firing and then hiring again.

Also it is so bad for morale that it doesn't even make sense from a business perspective, even if you can cut wages this way.

~~I think it only makes sense if you are effectively spinning out a particular produce into its own thing.~~ <- Actually this is sort of a different case, because it is more about changing the company people work for rather than getting rid of them.

my company was on a big hiring spree right before so it's extra lame

I can get hiring freezes, but for software/tech people are the core resource by which companies generate their product.

My employer (or rather my government due to being public sector) is also talking about large scale headcount reductions. That said mine may be a different case due to being government funded.

Let's say not a lot of people are happy about this (me included, of course)

What would be cool is if Google employees went on strike to protest the decision. Imagine search being down for a day, with an error page showing it cannot be loaded due to the strike. That would be fun.

Strikes don't really seem to happen in tech though...

https://www.instagram.com/reel/CmwwX8VtQ96/?igshid=OGQ2MjdiOTE=

https://i.redd.it/5skftx9lt2da1.jpg

W3C rejects Google's Topics API meant to replace cookies with an equally privacy-issue-ridden platform, one of the few of Google's ideas the W3C didn't rubber-stamp. Google vows to continue with it: https://technewsspace.com/googles-plan-to-replace-cookies-with-its-own-topics-api-protocol-is-in-jeopardy-rejected-by-w3c/

What I never get with these personalised advertising things, isnwhy they don't just use the site your currently on as the basis for their personalisation. That seems to me to be the most up to date signal of what people are interested in, and can be done without any stateful tracking.

that is still a tracking happening

True, but only that someone sees an Ad on that website. And that would have to be tracked to give Ad payouts to that website's operator. The advertising company only needs what page the ad is embeded into, which can be statically encoded when the adverts is initially included. They don't need anything about who the person looking at that ad is.

so non personalized ads

I guess the idea is that the ads would be served by a third‐party to the website with ads on it, and would make use of the data on the very page with ads to determine which of its poll of ads to show. Whether or not you want to call it “personalized” is a matter of perspective, I think. Different users would be seeing the same kinds of ads on a page, so it wouldn’t be personalized in that regard. But for the third‐party ad service, they would be customized per page that has ads. I suppose it can be seen as a shallow personalization, where the ad provider only knows which page you’re currently on.

so relevant ads

Yes.

But not all of the web is static pages. What does the above mean if "the page" is your facebook timeline?

"personalized" ads based on what the algorithm thinks I want to see usually have nothing to do with what I'm actually looking up and trying to find.

compared to just, like you said, using static ads or ads that are actually related to the same category of information you're looking at and reading about...

they work for me

but I also try to give as much information as possible

The most relevant ads I see in my experience are static ads with no javascript at all.

With the exception of paid sponsorships in youtube videos. Those things are irrelevant as hell

But aside from that, other types of static ads that are part of the static content of what you're looking at, tend to be way more relevant in my experience than "smart" ads

With the one exception of youtube sponsorships.

||raiD sHaDow LegenDS||

I’ve been working on the JavaScript chess library I created for my Lichess bot! I started working on adding support for more variants to it! I wanted for people to be able to set up their own games and chess variants using the same mechanism I use to define the rules of chess itself, so that it’s very general. The whole thing is admitedly fairly slow, but it works pretty well for my purposes! Now, the API changed somewhat signifcantly because of my changes. It’s a bit more complicated now, but I hope it’s not a big enough downside. I still haven’t updated the documentation for it, so I created a different branch for it for now, but I’d love to know if anyone has any thoughts to share about it! See here: <https://github.com/zamfofex/dummyette/compare/main...variants> (specially the new `variants.js` and `variants/chess.js` modules). It’s a fairly big change, so I’m still a bit conflicted about whether I should go through with it. Here is a slightly more thorough explanation: > Before, the “board” abstraction encoded both the state and game logic. But now, I separated into various different concepts: > - A “geometry” represents the shape of the board. Concretely, it is a set of positions (and by “position”, I mean like coordinates, rather than a chess position). > - A “storage” represents a board with pieces in it . Concretely, it is a map (a.k.a. dictionary) of positions of a geometry to either pieces or empty squares. > - A “state” encodes extra information about a chess position, such as castling possibilities, and whether e.p. is possible. Alongside a storage, they represent a chess position. > - A “rule set” encodes the rules for games and variants. Each of them can operate only on specific geometries that depend on the game they represent. Concretely, they have a method that takes a storage and state (where the storage is of a specific geometry type), and produces moves, as well as a method that validates a storage and a state by returning a boolean. > - A “move” is effecitvely just a function to storage and state deltas. A state delta simply tells “set the keys of the state to these new values”, and a storage delta is an array of movements. A movement tells “take the piece in a given square, and put it in this other given square”. > - A “board” is a snapshot of an ongoing game. Concretely, it just ties a storage, a state, and a rule set together into a single data structure.

> specially the new variants.js and variants/chess.js modules Like the actual idiot that I am, I had forgotten to commit those modules! 🙃 But they should be there now.

Google employees can't strike if they're not unionized

You don't need a union to organize a voluntary walkout among coworkers.

You only need a union to force your coworkers to walkout with you regardless of how they feel.

And to force everyone to pay union dues of course :)

you actually do

if everyone calls in sick on the same day it's called an illegal strike

I suggest you read the national labor relations act

and the laws in the state of california where google is incorporated

I thought there were already laws to protect workers from their employers if they do such things. This is the first time I am ever hearing of "illegal strike"

yea, not all strikes are lawful in the united states

it varies state by state

Maybe if there already is a union it's illegal to do that outside of the union

well for example, in Ohio, you must give 10 days notice before going on strike if everyone calls in sick, that's an unlawful strike

because there wasn't 10 days notice

do note that an "unlawful strike" doesn't actually mean you are going to be arrested, but it means you lose your legal protected rights as workers

Or maybe if you work for the government... Well I guess Google is government because they are technically God too

if you work for the government as part of a union the term for that is a public-sector union

examples of those would be first responders, teachers, nurses for public hospitals, etc.

there's certain laws that protect workers when a strike occurs. for example it's illegal to fire all the striking workers, unless the strike in unlawful, in which case they lose their legal protections

one of those laws is the national labor relations act

it's a pretty big deal, it makes things like pay secrecy illegal

and various other ugly labor practices

there's a lot of statute about this

Yeah... But if you work for the public sector then you usually, understandably, have less rights to go on strike or to unionize because you'd be unionizing against the government and the taxpayers.

it depends on what the statute says

and varies state by state

it's not as clear cut as that

There's usually a lot stricter rules if you choose to work for the government.

that's actually not the case

plus, a lot of the conditions are governed by the contract language in the labor contract

"what does the contract language say?" is a big question

If you ask me... You are better off not associating with Google.

considering that they just had mass layoffs I don't believe they're hiring

I know it is arrogant of me to say this since Google offers very good incentives to their employees. But if someone gets fired by Google, maybe look at it as a blessing.

Google does not deserve to have such a big part in our lives.

They are dangerous.

Everyone who works for Google makes them more powerful in this world.

And the more anyone uses or depends upon Google, the more power they have as well.

But that's just my 2 cents

quite liking this

new wiki is shit

you can always use a custom skin if you want

I hate This arguments

I can always write my new browser that will convert all the sites in the world into what I like more

but the default site is shit

I mean you don't need to, say, use CSS override or other mildly annoying things to do that

MediaWiki always supported user-customizable skins

I googled it, and I saw that in order to return the old style, I need a Registration

besides from that, I'm generally supportive of the width reduction because in mobiles you _will_ have to adapt to a narrow screen (you can't change the mobile form factor anyway) and yet the old layout will make most users unaware of this

`useskin=` parameter is a thing, though I'm less sure if this persists well throughout the session

oh unfortunately that doesn't persist

hmm, I think it should persist

wikipedia already has a mobile view. And it's terrible, always when I was on my phone I manually switched to the PC view

it might be that you just want the maximum amount of text regardless of the platform?

the question is different - why the should I agree that changing the design of the all site is for mobile phones

personally I already zoomed WP in the desktop so that its intrinsic width roughly matches my window width anyway, because small texts hurt my eyes

just use one of these they have the same goal but are coded differently https://greasyfork.org/en/scripts/458501-vector-layout-for-wikipedia https://greasyfork.org/en/scripts/458494-old-wikipedia-layout

I hate space, just useless space

how about zooming in like (the past) me?

well, great. I know that I am not alone and that there are already thousands of scripts

yeah those will simply put `?useskin` to every WP URL, shame it doesn't persist by its own

text size

uh, so you like small texts _and_ hate space?

I use this font because it takes up the entire size of the font space

symbol space

small texts ?

that's interesting to me, for me (and I believe it's a generally agreed wisdom in typography) contrast is much important than the actual space used or unused

it is not small, it consumes space more efficiently, because of this 1. it is better visible, and you can make the font size larger 2. it is clearly visible at the same size at a more distance

contrast? the text is black, the background is white, what kind of contrast are u talking about?

for example I can't easily tell a difference between `e` and `s` from your screenshot

(even when zoomed in enough)

you can't judge it by the first glance. It takes a week just to get used to it font

this font allows me to see everything without glasses. precisely because all its letters are discrete and have significant differences

I too have designed a bitmap font, and vertical segments are something you will always end up being suboptimal no matter what you do

anyway this shark disagreement between us is really interesting to me

👀

it might be also something to do with me having used dense-by-default scripts throughout my life

e.g. you can't really see a difference between "홋" and "훗" in most hangul fonts

so I think I'm naturally more worried about those fonts

This is a good example of how not to "improve" the design of a website <:KekDog:805390049033191445>

So much lost space XD

🤮

FYI my browser window is 1680 pixels wide and I always used 140% or 150% zoom for wikipedia, so the effective width would be somewhere around 1160 pixels

"The update initially received backlash, most notably when the Swahili Wikipedia unanimously voted to revert the changes.[80][83]" XD

at that window width I can't really use the website without zooming

oh, they have ported the silly new placement of “Languages” from the French Wikipedia to the other languages?

at least it’s consistent now, then

it was so silly to have in a different place the very thing that would lead you to the wikipedias where it’s in a different place

the button on the bottom right of the page makes it bigger although it's not persistent

come back

windows 11's fluent design is a definite improvement over its previous ones

i guess mainly because metro and mdl2 was just bad

I'm going to provide a contrasting opinion, and say that I prefer the new Wikipedia design. I find the limited column width much more preferably for reading large amounts of text.

With reddit I also slightly prefer the readability of the new one, however the old one didn't really have the same problem as Wikipedia because there weren't the large pages of text. On the other hand the performance of new reddit is quite terrible. They could really do with doing some server side rendering.

😂 👉 windows 11

Windows 10 isn't as bad as I thought

As long you don't use 21 inch display in full HD

On my old computer is fine

Same as Windows 7

Bluetooth works

Update it did all

My Windows 10 did all updates silently on 17th January 202

ehh, reading super wide columns is uncomfy anyway

I reverted to the old look, cause the new look is terrible

If you are a developer the columns are two

Hitori

what

You are logged on Wikipedia so you will see an even bigger vertical bar at right with the correllated edit and info for the Page views

Maybe it disturbs you

But I don't understand how one single vertical bar can bother

I'm against freedom

Or two or nothing

https://en.wikipedia.org/wiki/Wicked_problem

https://www.youtube.com/watch?v=IVfpIg24Kwo

Recently, 44 gigabytes of Yandex sources were Thief and Dropped to the network. and found this function. and the meme is that no one can rewrite this code in 3 days so that it runs faster and is not a table from else.

1976

well, the astronomers are well done of course, well, they're just handsome, yeah, well, can't say anything, well

did anyone tell them about jpeg xl?

No seriously, it's infuriating

https://www.youtube.com/watch?v=3uakz3yaQz8&list=OLAK5uy_nbBXA-zDYPMy5cgivS78XjQKb6ttksvBg&index=8

Annoying how cryptospammers can join a server, put spam in every single channel, I can conveniently ban the spammer and remove all their messages in a single action, but then it still looks like I have unread messages in all channels while there are none (the only new message was the spam)

yeah these indicators suck

blocking spammer doesn't clear unread counters either

When I saw it, I thought I'd accidentally navigated to the mobile version--I think that's all the new one is, it's just the mobile version on desktop.

same

but oh well its kinda readable even if different

once you get used to it

https://www.youtube.com/watch?v=JK_hBk2f01k

Hello. I have no real knowledge of how all the black magic in image compression works but I like testing with libjxl. I don't ask for implementation but is simply curious, why isn't TIFF a supported format to convert from? It's the most normal format to have export to when using no lossy compression in my experience.

TIFF is a container format that can store images in many ways

I haven't seen any tiff implementation yet that can correctly show all kinds of tiff files

There are subsets like baseline tiff 6.0 that aren't too hard to implement (though even that is kind of annoying), but making a complete tiff decoder is a lot of work

Writing a valid tiff file is easy, but reading all valid tiff files is kind of impossible and a moving target, since tiff is not a standard but it is similar to PSD: the spec is basically "whatever Adobe decides to do with it", which potentially changes with every update of Adobe software

E.g. a few years ago suddenly I saw tiff files in the wild with 24-bit float samples, which is not something most tiff decoders will handle but Adobe just started doing it anyway

One thing that is kind of silly in the tiff design is that whenever there are two ways of doing something, instead of making a design decision and pick one, tiff will just add a header field and have both ways. Like big endian vs little endian: this doesn't really matter, but tiff allows you to do both (which is nice for an encoder since you can do whatever you like, but for a decoder it means you need to implement both ways).

shifts impl difficulty to decoders

Endianness, planar vs interleaved, full planes or tiles or stripes, tiff can do it all

Even conventions like 0 = black, 255 = white are things you can choose in tiff, and if you want the reversed meaning ('chocolate' instead of 'vanilla', so 0 is white and 255 is black), that's also an option.

These are unnecessary complications imo. It would make way more sense to me to choose some normalized representation and force encoders to use that representation, rather than forcing decoders to handle all possible ways of doing things.

Anyway, these are some reasons why I don't particularly like tiff. But if someone wants to add an optional libtiff dependency to cjxl/djxl and add tiff support (or rather, partial tiff support, since of course libtiff is not complete), I wouldn't mind.

For CMYK this could be useful, since we don't have an input/output format for that yet...

Thanks for the deep explanation, does indeed sound like a nightmare to support. I've always thought TIFF was one of the most compatible and safe formats for long term storage

Just a word of warning that even knowing when transparency can be supported and when it really cannot isn’t exactly straightforward: https://github.com/guardian/grid/issues/2395#issuecomment-939602668. Hahaha. Ha. Huh?

Adobe and alpha channels is one of the biggest messes of total confusion ever. It's a total disaster.

Cue the lolz: https://github.com/gco/xee/blob/master/XeePhotoshopLoader.m#L108

PSD is a mess. I still don't really understand when an image is supposed to be rendered with alpha or not. You can have RGBA images in PSD that are nevertheless supposed to be shown as RGB, and they use arcane conventions to signal such things, stuff like setting the layer count to a negative number while setting the channel index to -2 to indicate it is alpha transparency as opposed to index -1 which is a selection mask, unless some new chunk is present and then it might be alpha too, but inverted and premultiplied while otherwise it is not inverted and only premultiplied if it is CMYK

(making stuff up now but it is stuff like that)

In Photoshop, they have alpha, opacity, transparency, solidity, masks, and clipping paths. All different ways of basically having pixels that are visible or not (or partially).

You forgot some 😆 . But here, as a user, I would defend most as just different tools (not caring how are they expressed internally).

Different tools is fine, but for an interchange format, it's good to have some kind of normalization and not have 10 different concepts to represent basically the same thing.

In a way, I like to think of jxl as a standardized and conceptually cleaner version of psd/tiff: it can also do alpha-blended layers, cmyk, spot colors, etc, but it is kept simple enough and it is standardized once and for all. Not like PSD which is a completely moving target, adding new layer fx and funky blend modes and all kinds of new stuff every year.

Yeah. But, I suppose, the internals of PSD weren’t really designed for interoperability. They would need to opensource the renderer. Some “layered” stuff can be manipulated in eg. ImageMagick. But even knowing if the file at hand contains stuff one couldn’t manipulate isn’t easy. At least for PSD there is enough information for the flattened composite (and if it’s there). For exotic Adobe TIFFs, even that is missing. At least this was the outcome of some weekends I spent looking into that which produced the funny graph in the linked issue above.

Yeah the thing is there is no interoperable layered image format atm. PSD and TIFF are a hairy mess owned by Adobe, then there's Gimp's XCF and similar things for other editors, but those are never designed to be interoperable, they're just the save formats of the specific software.

So here, totally unrelatedly, I wondered once if that’s not actually one of the reasons “web people” might be worried: bloated JXL for authoring flying around on webpages… and thinking they need to shield us from that. Not that one cannot embed 10MB metadata + 2MB CMYK profile into a JPEG…

Yeah I have heard some "web people" say such things... Not only authoring jxls, but also "too high fidelity for the web"

But that's something you can do in any format, avif supports lossless 12 bit but I wouldn't worry that the web will accidentally get bloated with those

With the second… I wouldn’t even know how (and what for) to argue. (I would always say I am the one who cares about lower quality than even they are thinking of, though!) But with the first… Tricky. My naive mind would see a flag in decoder that can download JXL only up to a point when the “composite” view of the image is available (without any “authoring-related”) stuff. And another flag, for web web authoring apps, that would say: gimme that first, but later gimme all the rest too. But no idea if that even makes any sense.

I can see that being an issue… if JXL gets **really** popular for proper authoring.

I mean, currently people use high res jpeg, psd, png or tiff as 'source of truth' images and then an image cdn like Cloudinary downscales it to various resolutions and encodes it to 'web quality'

If jxl becomes popular for authoring, it would be a big jxl as 'source of truth' image and small jxls for delivery (possibly fewer different resolutions than now, if we can figure out a way to deliver 1x and 2x effectively from the same file)

No idea what percentage of images on the web use image CDNs…

Probably a small percentage of total images, but a large percentage of views.

Depends a bit on how you define "image cdns". In its most general sense, it's any automated process to deal with image downscaling/compression. Then I think a big percentage of websites uses that: e.g. WordPress and many other popular platforms do come with that. Basically any website that is not some static hand-written page will likely use some kind of image automation (whether it's provided by the CMS, some home brew self-hosted solution, or an external solution). In the more narrow sense (external service, images served separately from origin host), it will be a smaller percentage but I would assume still a quite significant percentage of the high-traffic websites. Generally the very large ones don't use an external image cdn but they do their own thing (e.g. facebook, google photos, etc). Then there are big ones that use an image cdn but they tend to get custom hostnames so it looks like their images come from their own domain (even when they're e.g. served by cloudinary). Then there are smaller ones that visibly use image cdns (e.g. images served from res.cloudinary.com). And then there's a long tail of low-traffic, mostly static pages where automation is not as critical and bandwidth may not be a big concern.

Me: **tries to decide if I want to get an LG 32UN650** Specs on most website: "it has a 10-bit panel" "it has an 8-bit FRC pannel" "it can show 1 billion colors" <:ReeCat:806087208678588437>

Worst thing is... it's not just a case where 8-bit + FRC always gets advertised as 10-bit on certain monitors, because other monitors do actually show "8-bit + FRC" in the specs <:ReeCat:806087208678588437>

yeah it's unreasonably painful & that's why it always takes longer to pick things to buy for me than it should

displayspecifications.com is kinda good site for this stuff

Yesn't I guess

It says 8b FRC

Which is what half the websites say

is FRC instead of "real" 10bit actually visibly bads?

It's not really bad, but also not as good as true 10b

is there like any at all display w at least 4k & proper 10bit for 400~500euros

ones i stumbled upon were 8bit+frc

rip

i got secondhand Pioneer Kuro PDP-LX5090 & it's pretty epic

if using wayland it should work fine in theory

ive been using swaywm in 10bit mode for a while now & everything i threw at it worked good

X11 is different beast here though

i can relate with you lol, firefox still crashes semi-regularly here

just too used to good latency at this point

any compositor ontop of X11 just feels wooby, and there is no transparency stuffs without compositor >_<

not to mention 10bit and VFR things just working on swaywm

Over the many years have been studying different devices and rules of languages, I am increasingly convinced that English is well, an excellent choice from what it was

very generally speaking, no, because you have no restriction over A and B

of course the goal of A (watermark) is to remain subtly visible in C, so that's what a ML model or anything else can exploit

on the 20GB storage: if you do have that limitation you should have a very good understanding of whatever is in your storage

AI methods can quite easily get rid of such regular watermarks

most modern softwares are not generally designed for such cases at all, maybe they can cope with 100GB but 20GB is very severe

I implemented this at some point: https://cloudinary.com/product_updates/removal_resistant_watermarks#gsc.tab=0

Makes the watermark a bit wobbly (geometrically warping it) so it's different on every image, making it a lot harder for AI to find the pattern and auto remove it