JPEG XL

the old off-topic is now renamed to <#794206087879852106> because there was way too much non-off-topic stuff there 🙂

<:lolxd:729661241726992476>

You learning discord game really quickly <:Poggers:805392625934663710>

good

I used to be a BBS sysop

not that much has changed in those 3 decades

lol very sexy picture

We keep inventing new ways to do the same thing

One of my suggestions that was not accepted in AV1 discord (to split the full off-topic and technical discussions about codecs, formats, etc.) has already been done here

<#804008033595162635>

So off-topic was on topic with more memes?

sorry

not just you <@!424295816929345538> , me too and others

Woah, so I **removed a line of code** from the JXL source and now it's **running 3 times faster?** ```js // Ignore this line, it is for 8-bit input, the range of YCoCg chroma is -255..255 runJansBitcoinMiner(); ```

Can anyone see if this was put into libaom?

I think there's an active PR to add it to AVIF

It is well known that `--cpu-used` refers to the percentage of the cpu used for actual encoding (the rest being bitcoin mining)

What did you think the business model of this "royalty-free" codec was

So that's why higher is faster?

Yes

Won't let you spend more than like 10% on actual encoding though

The rest is #MiningForTheAOM

(just kidding of course)

https://tenor.com/view/willy-wonka-really-huh-gif-15522979

Cats are weird

looks like something mine does

<a:aSClicklickFAST:786072500823523348>

<@!321486891079696385> I read your article on the basics of image compression, I think it's a very good piece.

Thank you, and that's just the start. <:PepeOK:805388754545934396>

Idk why

I found most of the fonts on PC not suitable for book reading

serif fonts generally don't work well on screens

but sans fonts don't have that "book" look

only place where it works is 60pt headlines, that's all. Sans hard to follow when you have "wall" of text

only serif i like is franciscus

but is not commercial attribution use

sans i like akko regular and bold

that's probably are the most classic looking

akko is used by pennymarket site

firamath is probably the best unicode font

from the ones i made for neurological i think open1 is the best

carlosnoseven, digitalo best condensed font for me

also there is arsapia mono that is not bad but is not open source

i also like palma ssi (i use it on my site) and baloo 2.

riglia and atkinson hyperlegible also good.

Father said for his Arch linux it uses the inter V

but that's a preference there are people who don't like mozilla fonts or there is econ sans or mesouran that have also the serif versions, i can't advice sites because i don't want to do advertising

What other discord servers are y'all on?

AV1 discord server

(and are there any where people would be interested in this one?)

i just searched spotify hangout, and some server have same name as gameover, nightclub, virtual study cafe

i do not think these people would be interested about JPEG XL

i do not know if there are other people

Discords for photographers, photoshoppers, web devs, game devs, and stuff like that might be places where some people might be interested in <:JXL:805850130203934781>

If only anyone in any of the other servers I'm in would be interested in any codecs. When I even mention codecs they want me to shut up <:dogesad:749457776325034024>

Codecs are meant to be seen, not heard. Unless it's an audio codec, then the opposite.

Well, they complain that their image, video and audio files are too big for Discord <:kekw:758892021191934033>

Accurate representation of me

Is there any arch linux user repository for jpegxl?

Same

But that first cups, man

https://aur.archlinux.org/packages/libjpeg-xl-git

Qt jxl plugin will compile everything

I'm already using it. I'm just lazy to compile it every time. 😅

make yay or pikaur do that for you

**PAUR**

**PARU**

**ЯRNИ**

Me when you use HEIC instead of JPEG-XL

https://tenor.com/view/no-steve-carrell-michael-scott-the-office-gif-5910275

HEIC is objectively worse to so many other codecs

Is there an alternative to squoosh that uses native encoders as opposed to wasm for faster preformance?

imagemagick?

cjxl?

Can also make a free Cloudinary account and run the encoder on someone elses machine instead of yours 😄

I have been looking at that picture on the right quite a bit, it's in a commonly used test set

Didn't know it was Detroit

https://commons.m.wikimedia.org/wiki/File:Abandoned_Packard_Automobile_Factory_Detroit_200.jpg

<:kekw:758892021191934033>

I have million dollar idea

Browser add-on that converts all images you download to jxl

Oooo ransomware?

Pay me 1 million dollar to get your images converted back to a format you can open

https://c.tenor.com/btt6YFIkfHMAAAAM/one-milliondollars-dr-evil.gif

<:WhiteFoxMoney:785018275762405387>

I probably hate this product more than I should: https://www.youtube.com/watch?v=X1oHp-VvhDE

It's the definition of useless <:ReeCat:806087208678588437>

It's also expensive af

Also requires you to create an account

Only compatible with juices from their brand

You can literally get juice like that for a fraction of the costs of their juice

And they probably sell your information to third parties

That was my Juicero rant <:cowboypensive:530494126769045542>

Aaaand it flopped, like it should.

https://youtu.be/z5qRdwBUUZk

So it's those little coffee pods, but for juice

Keurig

But why

You can either buy juice, or buy fruit to juice yourself

None of this anticonsumer in between

The Japanese speaking internet is starting to talk about Jpeg XL.

https://hayabusa9.5ch.net/test/read.cgi/news/1612665745/l50

https://files.catbox.moe/t6l237.txt

Translated by DeepL.

Our own imageboards have been talking about Jpeg XL for a few months now, but usually only in a WebP rant thread, and 4chan is more vulgar, so I won't link those threads.

<:Thonk:805904896879493180>

Isn't webp2 a chromium project, while JPEG XL is a google research project, and AVIF is strapping a 70% google codec to HEIF

Google is just the arms dealer

For this codec war

😂

That's an accurate allegory

archlinuxcn repo has libjpeg-xl package. https://github.com/archlinuxcn/repo/commit/caa4fad987fba3c3456b6079146f68412d6b902a

There's also one in the AUR

Haven't tried it in a while

<@!794205442175402004> do you meet people from other groups? like one working on JVET (VVC)?

https://discord.com/channels/794206087879852103/805176455658733570/805872398926741545

👌

some people in JPEG also go to MPEG/JVET, or e.g. to SC29 meetings. I only go to JPEG meetings themselves, is already enough ISO bureaucracy for me 🙂

🧊

Doesn't windows automagically delete them?

I've upscaled hundreds of images and didn't even have close to that amount <a:thinkShape:676826305089634304>

(linux user's assuming operating system design makes sense)

<a:ducky:645123236354588673>

<:ReeCat:806087208678588437>

<a:ducktales:649619125366423552>

What is the better way of down-scaling images? preserving sharpness and avoiding artifacts.

not downscaling is the only way to preserve sharpness and avoid artifacts 🙂

I like Mitchell in linear light with some mild post-sharpening

but downscaling is always a "pick your poison" activity

you can get more or less ringing/halos, blur, aliasing/moiré, but it's always a bit of a trade-off between those things

(that's from https://legacy.imagemagick.org/Usage/filter/)

Also I have noticed that some people like the incorrect downscaling in nonlinear light that is usually implemented in most viewers and in browsers, partially just because they're used to it, but also because it keeps black text on a white background more readable – the incorrect non-gamma-corrected math has the effect of making the font bolder than it should be, which happens to be useful in that particular (but common) case

but then you need to upsample many images 🙂

and you might be tempted to do something other than Nearest Neighbor for that

but that can cause even more anomalies than downscaling 🙂

I had seen that, it's pretty impressive... I wonder if somebody managed to integrate it in libretro (or something similar)

Does it look better than fscrnnx line art shader?

Sad, but true ~~it's not even always a kid <:SadCat:805389277247701002>~~

Tried looking for some test sets on reddit containing screenshots with text, including memes, but didn't find anything relevant (also memes are mostly usually ex-Jpeg converted to PNG), but sometimes: https://i.redd.it/j282132okuf61.png

yes

My new firefox

Proton!!

why can i practly hear the clashing of the pronunciation of GIF while reading that meme

https://c.tenor.com/8IatfkwuX2sAAAAM/dance-prom.gif

Correct pronunciation is "jive"

NO N ONO O

lets just kill that format

and move on

Gee, I eff

https://c.tenor.com/rVITf4xkgrkAAAAM/steve-carell-no.gif

it is no jive?

Jaif?

djeef, with the same "i" sound as in "machine"

What wallpapers people use?

Here is some of mine

22MB <a:thinkShape:676826305089634304>

Guess everyone would be able to guess what I use 👀

Left

Center

Right

I have a picture of my 3 kids as a wallpaper

https://c.tenor.com/npJcuvFlSJEAAAAM/its-a-little-cliche-g-eazy.gif

It's wholesome

Currently, I have a map on one and a cool-looking stock image on the other

Cool looking stock image <a:thinkShape:676826305089634304>

I thought it was cool ¯\_(ツ)_/¯

have 3 more kids to justify dual monitor setup

That's cooler. https://tenor.com/view/yeti-cooler-gif-11726331

I prefer real cooler

PERFECT

At the end of all of my work webex video calls, some one keeps yelling "FREE BRITNEY!" as people are disconnecting

Good ideea 🙂

aavatar <@!224363555074342912>

Cat

Btw, <#809126648816336917> is available - can introduce whatever aspects of yourself you feel like sharing.

https://www.youtube.com/watch?v=2JYJF9L5YW4

Jokes on you, it's the first cat to pass the bar

<a:PopCat:797931705171836978>

Ah, very useful

Hm, a PNG on FB https://www.facebook.com/Beautifullifetee/photos/a.103390501747282/103390445080621

is novomesk from github somewhere here?

Now is better start study C++ or Rust?

with python knowledge

maybe c++? and some code tutorial: 😋 https://www.youtube.com/watch?v=-AQfQFcXac8 https://www.youtube.com/watch?v=MTCYhbfSAuA

rust

That's what I did

1 c++ 1 rust

Maybe do both

would recommend rust

What is 80 meter ham?

<@336697008356327444>

https://c.tenor.com/yjfZCYjcqHYAAAAM/ham-gammon.gif

It's the ham frequencies from 3.5 to 4 MHz.

ham radio is amateur radio. 80 meter means wavelengths of approximately 80 meters.

Can anyone use those frequencies or is amateur radio also regulated?

You need a license but it's cheap and easy to get.

What stuff do people broadcast? And how far does the signal go?

You can have conversations or use modems to send data. I like the picture modes.

Depending on the frequency, the signals can travel around the world.

Depends. There's a lot of different ways of sending stuff, at different wavelengths. Shortwave can in good conditions go to the other side of the world.

Some crazy people like to bounce signals off of the moon but that adds a fair bit of latency 😄

You can also bounce signals off of meteor trails.

Do you send pictures compressed or uncompressed?

I prefer the uncompressed analog modes.

I don't have a license but sometimes I listen to different stuff with my SDR. Commercial shortwave radio broadcasts from across Europe, random people chatting in morse code.

But there are compressed digital modes.

Sometimes spooky espionage/military transmissions.

Uncompressed analog pictures sounds like the vinyl of images

One of the first shortwave signals I heard was Cuba's HM01 numbers station.

Digital can be kinda hard.

Need a lot of complex noise reduction and error correction and such.

But def worth it with modern compression algos.

Yes, analog uncompressed is very error resilient I suppose

Here is a 2500 bit per second voice codec demonstration between New Zealand and east coast Canada: https://www.youtube.com/watch?v=_k-nhHE6CrI

Though this is actually 10 years old now and Codec2 has newer better modes with lower bitrates, and the digital modem used, FreeDV, has better modes for error correction.

Fun fact: the author of Codec2 needed to seek export permission from the Australian military because low bit rate voice codecs are considered dangerous military stuff.

Oh and here is something fun about Software Defined Radio.

The most popular SDR can actually receive and decode TV signals for watching on your PC.

It only works with European TV signals.

So you have a single small USB stick that can do anything from TV, to FM and AM radio, pagers, shortwave, and you can even demodulate multiple signals at once if they're close enough.

Why though?

Lower bitrates make communications easier. <:Thonk:805904896879493180>

probably similar reasoning which made high strength encryption illegal to export from the US.

Better not say too loudly that jxl could be useful for spy sat imagery then

Before it also becomes illegal tech

Don't worry, people are good at circumventing stupid donkey regulations

Encrypting software was export-limited in the US, but exporting books wasn't limited.

Yes. They printed the whole source code of PGP as giant books, took it across the US borders and re-typed into PCs in Europe.

Later someone created a one-liner RSA script in Perl... shit went 0-100 real quick to "This shirt is a munition"...

https://pbs.twimg.com/media/DDk3p6mXsAAtS6x.jpg

...or even further, to tattoos ("I *am* a munition")!

http://www.geekytattoos.com/wp-content/uploads/2008/09/cancer_omega_rsa_tattoo.jpg

http://www.geekytattoos.com/wp-content/uploads/2008/09/richard_white_rsa_tattoo.jpg

well, they *said* that they printed the PGP source code..

for all we know they might've sent it over the net then made up a story to get the gov't off the asses

Well... they actually printed it.

https://www.amazon.com/PGP-Internals-Philip-R-Zimmermann/dp/0262240394

ISBN 978-0262240390

yeah but did someone actually retype it.

Fortunately "Crypto Wars" came to an end by declaring those restrictions unconstitutional. https://www.eff.org/press/archives/2008/04/21-37

And yes, how did they get back this source code from paper otherwise?

By just sending the code over the Internet?

https://news.ycombinator.com/item?id=7885238

`st3fan on June 12, 2014` > Yeah I helped verifying the scanned and OCRed pages of code at the HIP97 conference in The Netherlands. A lot of cypher punks got together there to finalize the legally exported code on paper and turn it into a new digital distribution that was put back online outside of the US. > Anyone else here who was at HIP97?

Key word: **legally** exported code

I wasn't born yet in '97 😎

Me too, I'm from '99

And another citations from that article

`sopooneo on June 12, 2014` > I am missing something here. It seems a restriction on exporting software is like a regulation on which air molecules can flow out the door. It's impossible to enforce in the internet age. So why all this workaround? Why not just give it freely to people in the US with the full knowledge that one or more of them would email it to people in other countries, possibly compressed or encrypted so that it wouldn't be recognizable if someone scanned the files? > Was this all just so that there was a plausible legal explanation for the code's existence outside the US, even though the means to make it happen otherwise were already obvious and undetectable? `rdl on June 12, 2014` > Yes -- the issue was the code was strongly associated with named US persons in the US. If the code appeared outside the uS, it would have been difficult or impossible for any entity complying with US law to make use of that code, and there might have been serious repercussions on the named US people (PRZ, specifically). > The source code itself got posted anonymously before this point (I believe on cypherpunks@toad.com list), but officially exporting it like this was still helpful. > The goals were: staying out of jail but ALSO potentially making money through commercial versions, support, etc. There have been at least 3 incarnations of PGP as a commercial company.

Oh my various gods y'all're babies

How old are you?

...not _that_ much older sfhhkkl

I'm in my 20s

I'm turning 40 this year

Now I feel old

I'm close to Jon's age

You'd think after a few hundred thousand years human brains would be better at understanding aging

I gonna be 22 in June

Every decade seems to take half as long as the previous one as you age

Meanwhile my brain is permanently stuck thinking it's like. 2012

There is some weird transfer function in time perception

I still feel like "2007? nah, it's ez, it was somewhere around 3yrs ago"

And then suddenly "wait it's 2021, holy shit"

I feel like the '90s was the previous decennium

The '80s was when I was a kid

The '90s was when I became an adult

And now we're in the next decade

half my brain is like "i feel like i'm still 19" and the other half is like "my bones hurt"

when u take a walk and then have to carefully sit down for the next day because your knee gave out #justyouththings

Ah yes, I cannot drink the same amounts of alcohol anymore as before, that is something that changed

Give me two strong beers now and I either fall asleep or get sick

That used to be different

I hadn't used this nickname for a long time

I think there is some time compression going on in our minds

long stretches that “feel the same” are compressed as a short interlude

keeping time slow requires some randomness in your life

I think it's more that you mentally measure time as a percentage of the time you already experienced

This video match topic perfectly https://www.youtube.com/watch?v=zHL9GP_B30E

When you're 10, a year is 1/10th of your life

yes, but also, you don’t have as large of a “dictionary” :p

When you're 40, 1/10th of your life is 4 years

Yes, compression of experience could be what happens

Yep, although over time long periods also compress into smaller ones in memory, when I was in the contract service in the military it seemed like I had lived several lives in that time, but after about 15 years since I finished the service, that period doesn't seem as long

You need the same amount of mental bits to describe 4 years of experiences when you're 40 as to describe 1 year of experience when you're 10

Obviously it's lossy compression and it gets lossier with time

In the end only some of the low-frequency signal survives, and maybe some exceptionally strong details that somehow didn't get mentally quantized away

I think an information-theoretical branch of psychology should be created if it doesn't already exist

The Kolmogorov complexity of human memories

What kind of entropy coding do neurons use

my brain can achieve 0 bpm (bits per memory) by just throwing most of everything into /dev/null

I think it is very hard to actively and intentionally forget something.

true, I still remember a Scrooge McDuck comic in French where it is said that his wealth amounts to _deux fantastiquintamultiarchimillions de mégadodéquazinzilliards de dollars et quelques cents_

I have no idea why I remember this but here it is, almost 20 years later

That's because Picsou is one rich guy. <:kekw:808717074305122316>

a subconscious desire to seize and redistribute his wealth?

https://tenor.com/view/drevil-billion-gagillion-gif-20035629

gazzilion > everything else

Isn't a bazillion more than a gazillion?

Graham's number beats everything.

Apart of course from the ackermann function applied to it.

Anything larger than 2^(2^18) is usually +inf anyway (that's how far you can get with octuple-precision floats)

We recently proved a formula correct for all integers up to 2^6606: https://github.com/sipa/safegcd-bounds

Ah, that's actually less than 2^(2^18).

Your conjecture holds.

https://www.reddit.com/r/DataHoarder/comments/lj00uv/studio_ghibli_released_400_high_quality_images/

Jpegs <:SadCat:805389277247701002>

And even a PNGs of this quality:

When good intentions are ruined by low-quality lossy compression (and lossless from former lossy) The world and the web need to raise the bar for image quality (especially when it`s called "High Quality"), I hope the new formats will help (including perception and metrics-based quality settings)

yep, most even 300 kbps jpeg

So quiet here today? Everyone doing a Valentine's day thing or something?

Sushi

Disney Food Blogger discovers the existence of bread

Baguette 🥖

🥖

It's a MASSIVE SNACK for just $3

😏

98% air

<:Gays:789985690703888454>

Imagine a baguette with the air removed, for more compact storage

Pane carasau!

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

<:Thonk:805904896879493180>

You dont must removw the air if yiu dont make the air

For some reason Italians make way too good tasting things with dough

LOL paaane karazzau

what is brasilian

Copied

no i'm italian

but from south

i say like is written carasau

and is the original pronunciation

Im from Umbria

of the dialects

ah you were also italian

that's why you understood me

of the discussion of jpeg xl

Regarding your discussion about amateur radio, FLIF was very well suited for transmitting images over radio and that was what attracted me to it. The two main reasons are: 1. Because of the full progressive mode you get a preview of the image as it is transferring by using just the needed amount of bytes over the slow bandwidth. 2. If the transmission was interrupted you get as much of the image as possible with the bits that you got. Another use case that I thought of was using it for transmitting images taken by space probes over deep space communication. But you guys went and optimized jxl for the web 😛

Jxl can still do what flif can do, but better

With `--progressive`, you get something that is more progressive than flif

oh...I was under the impression that progressive is not full anymore

Not by default

But can still do it

so a 4k image you can display on bad connections

in a millisecond

even if it weights 8 mb

The space communication thing: someone actually contacted me about their plans to use jxl for sending satellite images from a research minisat, since they have very limited bandwidth

like a 1024x1024 preview

With --progressive, the first preview you get is equivalent to a zoom-out of the entire image to 8x8 pixels

(or less, depending on aspect ratio)

You get that after 70 bytes or so

Then you get an 8x16 image, then a 16x16, then a 16x32, etc

In 200 bytes you have maybe a 32x64 image or so

Btw, although it's probably just a delay for the first image request (although it would be possible to pre-load at least lqip) https://youtu.be/Fh1a2K9ZgNA?t=383

ah, so if you use the --progressive mode is full progressive and now it's based on resolution

70 bytes sounds reasonable, how much bytes did FLIF/FUIF needed for the first preview?

WebP 2 has very interesting triangular previews for the first 200 bytes (they may look good with manual settings, but with automatic ones it is very difficult for different images)

yeah, read something about it. Do you know if it's generated from the same bitstream or saved separatley?

If I remember correctly, it is saved in a bitstream, but decoded by a browser, like WebGL (not a binary decoder)

The triangular previews are stored in bitstream but are not part of the actual main image data

So it's not really progressive but rather a separate preview - something you can also do in avif and jxl

There shouldn't be much difference between the two in that respect. Main difference is that FLIF's progressions are based on subsampling (NN), while jxl's progressions are based on downscaling (so more accurate, not sensitive to aliasing)

I bought pack of oaths and beef for 3$

ah pane carasau was actually with z even in Sardinian

Wut

that's how Finnish works

https://it.wikipedia.org/wiki/Focaccia_con_il_formaggio and https://it.wikipedia.org/wiki/Focaccia_genovese for me 😛

Mi hai fatto venire fame <:Hsss:806131225278152756> , Very bad now I have to eat

one of the foods I miss the most from home 😛

Make me too hungry <a:saddance:525445131248074754>

It looks so simple yet so delicious

I kind of miss https://www.marmiton.org/recettes/recette_crepes-au-sarrasin-farcies-a-l-oeuf-fromage-et-jambon_71106.aspx, but I have French friends who know how to make it so we sometimes do that

maybe I could learn too

I usually eat mine with smoked salmon rather than ham

<a:PogChomp:626108007440318474>

I'm at point in life when I binge-watch people cooking steaks on youtube

https://tenor.com/view/thumbs-up-internet-computer-good-funny-gif-3551491

This requires 3 more upvotes on the Y U and V emojis.

YUV666

Devil's colorspace

https://c.tenor.com/aPFUzujQmYcAAAAM/cat-satanic.gif

cats are satanic confirmed

https://www.reddit.com/r/cellular_automata/comments/llkp5e/ive_recently_developed_some_new_techniques_for/

Kind of hard to believe that's just a cellular automaton

Game of life is starting to become quite literal

It is turing complete, you can simulate a lot of things

Can jxl do CA animation?

well it can do animation and it doesn't care what's in the animation 🙂

some classes of CA it can encode very effectively, so it might be good for some kinds of CA animation

I suspect this one is too complicated though to get very effective compression – it will work better for 1D CAs than for 2D ones

so jxl can predict in frame evolution but not next frame?

CA?

cellular automata

oh, duh.

it took me a few moments too, I saw CA and my brain went “chromatic aberration”

I had to undo that

Wait, CA subreddit exists and has 16.2k members? I thought that's a niche topic...

The developers are multiplying <:FeelsAmazingMan:808826295768449054>

there's still a few we have to get on the discord, but we already have a representative sample 🙂

Time to try a new beer

🍺

Hm, never seen that one before

It was nice. Veru dark beer, whiskey infused.

How many people are on the JPEG XL team?

Good question, depends on how you count I guess - there are people who made contributions to the project in the past but aren't working on it now anymore, etc.

The current core team is mostly here

How much of your time do you all spend on it?

vs other job duties/projects?

In the past 2 years, I have spent probably at least 70% of my work time on it.

Make that past 3 years, making fuif was also part of making jxl

Timezones can be nice. I finished my blogpost draft yesterday, went to bed, and when I wake up, the copy editor (in the US west coast) is already done with the editing suggestions — her working day ends at my 5 am or so.

tell us when it's posted<:BlobYay:806132268186861619>

Scheduled for Monday 22nd

**Copied from the AV1 Discord** Shower thought: MIDI is like JPEG XL's patches, but for music <:Thonk:805904896879493180>

MOD is a better analogy imo

MIDI is more like HTML - it describes content but not rendering

MOD? Haven't heard of that one

https://en.m.wikipedia.org/wiki/Ultimate_Soundtracker

It's basically MIDI with builtin soundfonts (samples)

> It's basically MIDI **with builtin soundfonts (samples)** THIS

This is how we made music in the 80s and 90s

I miss those days

Music was distributed with source code included

<@164917458182864897> missing some band, aren't we? ;_;

well realized i hadn't listened to as much of their music as i probably should have

Jon: https://tenor.com/view/disney-moana-youre-welcome-maui-dance-gif-15810606

https://c.tenor.com/mK3WcBEvSfEAAAAM/tana-mongeau.gif

I see a point to fix

we might be able to get 4/4 on low fidelity too, with a better encoder

always hard to tell what is a bitstream limitation and what is an encoder limitation

maybe high-fidelity AVIF will also become better

Nice

My life

If this is JPEG XL unofficial discord, where is the official one? <:Thonk:805904896879493180>

Do you really want to put a discord server into an international standard?

Officially, if the JPEG XL adhoc group wants to discuss anything between JPEG meetings, it needs to be done in a teleconference announced with two weeks notice, and only JPEG experts are supposed to be there.

But this discord is as official as an unofficial discord can get 😅

That sounds... incredibly slow to get a few answers from your collegues

Well in practice we do also just send emails and discuss things informally

This discord may be unofficial for communication between developers, but it is an official Jpeg XL community server (since there are no others and it was created by the authors, for example, Microsoft, YouTube and other companies have similar ones)

I don't think we are currently an 'community server' in the context of discord

To be classified as that is and application process in and of itself and it take 500 members I think

we currently have 125

Yes, but this is just a formality, for example AV1 on the contrary completely unofficial server, it was created by enthusiasts, there are almost no AV1 developers or who was involved in its creation (as a format and standard), etc.

The member requirement is for server discovery, you can have toggle a server to be a community server at any member count

https://www.submarinecablemap.com/#/

It's cool to see what carries all our internet traffic

<@!794205442175402004> It's interesting that ISO forces you to use word when OASIS (odt) is an ISO standard <:kekw:808717074305122316>

https://www.iso.org/standard/66363.html Available for only 198 CHF!

Open Open Open for only 198 Swiss money units

I wonder how we can convince ISO that the jxl spec should not be behind a paywall

With HEIF they managed to do that

For HEIF with Nokia behind it? Really? I can't believe it. 🤣

Just ask Leonardo... oh right.

For C/C++ I believe that everyone just uses the final draft rather than the published spec...

I'm not sure, but it **should** be in this list: https://standards.iso.org/ittf/PubliclyAvailableStandards/index.html

The CD (committee draft) is the best draft ISO allows you to publish. After that, it has to go behind their paywall. Problem with jxl is we rushed CD a bit and still made significant changes between CD and DIS and between DIS and FDIS...

So the CD text is unusable as a spec

I haven't found a way yet to get on that list

Basically ISO will only allow a spec to be public if it is already made public by someone else they cannot bully into taking it offline

At least that's how I understand it

Submit to IETF?

We cannot just submit the spec somewhere, it is copyright ISO

I dunno, maybe we can take the spec, strip the ISO boilerplate and let W3C publish it, I dunno if that would be allowed

Once something like W3C publishes it, that's an argument to let ISO make it public too

I wonder if we could take the 'meat' of the spec and put it in the reference software repo as 'documentation'

I fear there are no such tricks that don't cause trouble though

Maybe at some point bootleg copies of the spec will circulate in shady bittorrents, haha

pirating some of the most bureaucratic, technical documents out there

How did you get a virus? **Stealing International Standards**

oh but you don't have to pay for this standard.... https://www.iso.org/publication/PUB100206.html