JPEG XL

No 73 123

Esme 112 61

Considerino is v2

90 63 this

Galmuri mono 9

56 79

There is another font called Averta but is a web font

https://www.cufonfonts.com/font/telka-trial

https://www.cufonfonts.com/font/highbrow

https://www.cufonfonts.com/font/spoof-trial

https://pbs.twimg.com/media/FzLPfItXoAMbjsN?format=jpg

https://www.cufonfonts.com/font/cartoon-network-2

Did you make that font <@416586441058025472> ?

No

*Enterprise moment*

https://youtu.be/tMvGGQZHMjc

https://mathstodon.xyz/@MoritzFirsching/110593663999544679

https://www.itu.int/rec/T-REC-T.24-199806-S/en

Found the source

It was CMYK originally, lol

If the picture was taken when the newspaper was still fresh, it took 10 years for it to become part of a published standard (from 1988 to 1998)

Must have been fun in 1988: Read the paper while having some wine, snails and lobster and then take the bike to play some tennis...

The life of an image compression researcher has changed a bit since then...

or cycling while eating a lobster !

Acrobat shows RGB, but knowing properties of an image within a PDF requires a PhD I do not posses, so may be wrong:

The PDF is a preview, it even says on the site not to use any images from it and to contact them for the real download

Yeah you can buy a CD-ROM, lol

looks like the embedded images in the PDF aren't the full versions

SyQuest for a small premium 😉

This is a real off-topic, but I thought someone here might know their VIPS… I’m trying to (help) switch the app from ImageMagick to VIPS and stumbled upon an issue. I want a command that will take any file and produce a PNG in sRGB (later I want it quanted to 8bit, but haven’t got there yet). This is simplified case: `vips icc_transform 16bitAlpha.png 16bitAlpha_out.png sRGB.icc --embedded` If I run it on 8bit input PNG: it works fine. If I run it on 16bit PNG, it produces one bit alpha. If I do `vips icc_transform 16bitAlpha.png 16bitAlpha_out.png sRGB.icc --embedded --depth 16` the 16bit works fine, but the 8bit is empty (sic!). (same thing happens when I feed VIPS TIFFs) Am I doing something basic wrong? (like not converting all the inputs to some common ground first: ImageMagick doesn’t require that). I thought I will ask here before making an idiot of myself in VIPS issues on github.

Here are (perfectly vanilla) input files:

Zstd is facebook, and facebook is positive about jpeg xl. So being against zstd is not really practical stance

technically speaking though zstd is not particularly suitable for web, making it indeed dubious for new inclusion

zstd shines when you have to compress really fast with a reasonable ratio (comparable or better than stock zlib), but main use case for content-encoding is decompression

brotli also has a built-in dictionary more or less tuned for web payload, while zstd doesn't (both support user-defined dictionaries though)

so feature-wise they are roughly identical with different implementation trade-offs

and zstd's trade-offs are not aligned with web while brotli's are

that said, is zstd really that large? I think without any compat supports it is slightly smaller than libjxl

especially when we are talking about decompression-only support

now, jxl is featurewise much richer than avif and can't be compared in this way

(just to clarify)

that said, I don't think zstd support will actively harm web platform, it's a nice addition to have but with not much additional value per se

zstd content-encoding will indeed make serving large files with comperssion on demand more efficient

given its ability to adapt the I/O rate and very good default options

mid-sized files won't benefit much from zstd (brotli is already fast enough for this use case)

and brotli will win for small files

that's what I say when talking about trade-offs

there have been discussions about using Brotli's dictionary with zstd on the web.

my local copy of libzstd is 702K and my copy of libjxl is 4.2M

Also, this is just adding zstd as a test behind a flag, we have no idea where zstd might go. They may well decide to throw it out after a few versions like with jxl, who knows.

i think because only the decoder needs to be counted, also at least chromium already has a lot of the dependencies needed for libjxl (so it's something like additional ~120K+ for jxl) as for speed, brotli is not as optimized as zstd and might be almost as fast, also as far as I know brotli works better for streaming data and decompression speed should be enough even for very fast networks, so for the web zstd has almost no advantage

Better to compare sizes with symbols stripped and libjxl_dec only (browser doesn't need an encoder)

Iirc jxl added less than 200 kB to Android Chrome (which is the one that cares most about binary size)

symbols were stripped already. I was just looking at what the distro I use ships

my own compiled copy of libjxl_dec after stripping is 2.1M

but as pointed out it may contain things that chrome already has

The x64 version is much bigger than the arm version because x64 has more SIMD variants

makes sense

is there a flag to disable SIMD?

Yes, or you could also compile just the best SIMD your cpu supports — but that's not practical for distributions other than Gentoo or something

indeed.

incidentally, I'm working with zstd compressed data right now. downloaded a copy of almost everything ever posted on reddit the other day, and the data is divided into files per month each compressed with zstd

You mean a full archive of Reddit? I'd have to assume text only

yes

compressed it's about 2TB for 2006 to 2022-12

Not bad

I wonder how far you could go down if you used lrzip with zpaq

something like two years, maybe

and then again when decompressing

zpaq strikes me as more useful for incremental backups than for just compressing a big file

I know I don’t chat here often, so I hate to come along and be, like “hey, look at this thing I made”, but I made a simple (and currently slightly incomplete) WebAssembly interpreter in Scratch, and I’m fairly happy about it! <https://scratch.mit.edu/projects/867209277> (If anyone has any thoughts to share about it, I’d love to know!) My plan is to (eventually) add `i64` support (and implement more of the WASI API) to run wcc (which is self‐hosting) on it so people can compile C and run programs within Scratch itself! The maybe (if it seems feasible) add floating point support to run Python on it too.

lmfao

nice work

i love absolutely cursed projects in scratch

I found a comment on the Chromium bug tracker saying that a size optimized and stripped decoder only build of libzstd is 99kB.

That's only half of libjxl-dec then

huh, the issue history is kinda funny. discussion in 2021 about zstd maybe being worth trying, and then absolutely nothing until just a couple of months ago

oh, right at the bottom is a comment from Jyrki saying that they should work on making Brotli faster instead, lol

i see only one useful way for adding zstd, if there will be a single unified brotli-zstd decoder, since they have similarities and because zstd as a library is more actively developed and optimized

https://youtu.be/PxT2MOsViBk

such an interesting cognitive glitch

Probably also has implications for subjective testing methodology, and could partially explain differences between results from pairwise comparisons (comparing two distorted images) and results from traditional MOS experiments (grading one distorted image on an absolute scale).

Bah, our CID22 paper was rejected. For one reviewer it was too hard to understand, for the other the dataset is not large and extensive enough. Sigh.

The problem with anonymous peer review is that there's little incentive for reviewers to do a decent review. It's more of a lottery game than a real merit-based selection mechanism imo. Then again I probably always say that when I get another paper rejected.

Basically you wait several months to then get a rejection based on two anonymous reviews by people who didn't really bother to read the paper but don't like it anyway

One reviewer saying there are two many references to related work, the other saying there is not enough related work mentioned. Ugh.

I suddenly remember why I left academia 🙂

<:FeelsSadMan:808221433243107338> at least the community respects the work put in

Where was it submitted?

Can you send it in again?

I generally feel that anonymous reviewers are either very on the spot or very off the spot

while my academic career was not very long I had a very helpful anonymous reviewer, but I've also received another anonymous reviewer who was clueless about our work (but felt it to be important enough to be accepted, hmm) for the same paper

one anonymous reviewer is enough to reject publication?

There were two reviews, both not good enough to recommend acceptance

yeah, that combined with acceptance rates makes or breaks your paper

these reviewers are kind of unreasonable, not much useful I can do with their comments

> The evaluation of image encoders in terms of compression and visual consistency, although commendable, does not significantly contribute to the existing body of knowledge. > > The presented dataset, CID22, originating from 250 pristine images, is not sufficiently comprehensive to justify the claims made in the paper. A more extensive and diverse dataset would enhance the reliability and generalizability of your results. > > The proposed new metric, SSIMULACRA 2, while intriguing, requires further validation and comparison with established metrics to demonstrate its superiority and utility in the field of image quality assessment.

so basically that reviewer is saying it's just not a big enough contribution to be worth publishing — nevermind that this is the first time the concept of "encoder consistency" is even examined in the literature and that the dataset is larger than any other iqa dataset described in the literature afaik, it's still "does not significantly contribute" and "not sufficiently comprehensive"

the other reviewer had these comments: > 1. The abstract should include relevant percentage improvement results to numerical measure the quality of some aspects of the work. > 2. What are the keywords? There are some problems in Keywords section. > 3. The structure of this article is rather confusing, and the logical relationship between paragraphs is not clear enough. I suspect whether the workload of the paper is sufficient. It is hoped that the authors can adjust the relevant content appropriately to make the paper more understandable and fluent to read. > 4. This paper is awfully written. There is no adequate objective and no reasonable conclusion. I have doubts about whether the theory/calculation are sufficient to arrive at a useful conclusion. > 5. The authors abbreviated so much that it was hard to understand their paper. > 6. The research problem and method of the paper lack innovation. The presentation needs to be significantly improved to bring out the abstract, the contribution, and make the problems, methods, experiments easy to comprehend. If they cannot then it is better to decline to submit.

basically this reviewer is saying "give me a better summary, I don't want to read papers"

So we have one who wants more data, and one who can't read data

https://lemmy.world/post/799816

Add that information prominently in the paper and resubmit <:kekw:808717074305122316>

Did you do the annotation of the 22153 images manually? How does annotation look like? Could community help there to maybe 10x the data?

the annotations were crowd-sourced, see https://cloudinary.com/labs/cid22

this was the paper we submitted: https://cloudinary-marketing-res.cloudinary.com/image/upload/v1682076683/CID22.pdf

we did about 46k small test sessions, I don't know how many unique individuals were involved but it must be something between 10k and 40k

Ah, so it wasn't the extended one that got submitted?

we paid about $1 per test session to get those opinions

a bit less

no, there's a word limit so had to shorten it, what we submitted is exactly at the word limit

I think it's kind of arrogant (or maybe just ignorant) for a reviewer to suggest that our dataset should be "more extensive and diverse". I think they don't understand how large it already is. If you want something significantly larger, you're going to need a pretty big budget, and even then you might run into the problem of limited humans — even crowdsourcing platforms do not have an infinite amount of humans to look at images for you...

(This amazing application restarted itself once again while I was typing...)

Yeah, it is either big budget or way less human resources (with the risk of more tendency/bias/errors)

lol

A person can look at 20k images within a day, that's no problem. But they won't grasp all the details and getting good annotations will likely take several weeks or months. If it was setting keywords on first impression or categorizing it with a very limited amount of keywords/categories then that won't take a long time however*

(* I did that once in a task for DHL — categorizing texts with only a few sentences sentences into four categories iirc — while watching dubbed anime, lol)

Published in a peer reviewed revue or not, you still contributed to the progress of science (since the data and the results are available) ... Is it so important to be published?

Nah, not really, imo.

But how does it look? 😮

i like

what is this font supposed to emulate? an 8-year-old's handwriting?

it's a little bit strange that some of the letters are perfect e.g. `o` and `r` and some letters are clearly handwritten e.g. `m`

That you're right

m and especially u

And the word ci

Are ugly

The release was rushed

I did only for me

Im not a designer

It has small x height as a feature

Even the e is worse

Is not being a replacement of commit Mono

I'm not a Bachelor student studying 4years

When there is more complex text it looks like a garbage

Readable text looks like this

I made my font condensed

I rewrote for condensed

Then I restore the design

Because it looked like a weird Galmuri 9

Mostra pulsanti

Real font should communicate button for pulsanti

This font is a bir too condensed

I don't know but I made it

It would seem like a strain in the eye

But i'm able to read all in this morning I woke up now so I have fresh eyes

The thing that I'm bored

i made a programmer font once

for pros

at gramming

I and f should be made a space

Only this would significantly improve the legibility

Also words like Teheran should display correctly

And for that you need to pay a designer

This is the amount of text is written on AV1community

But the same 20 old people

Hm, 1024 members <:thinkies:854271204411572236>

https://commons.m.wikimedia.org/wiki/File:Seitengr%C3%B6%C3%9Fe_PDF_7.svg

Maximum size of a pdf

We don't have dpi in the jxl header so I guess we don't really have a maximum physical dimension

What I hate about PDF is that it looks like documents aren't really future-proof

There was a audio feature but it used flash technology in most cases

It doesn't work within PDF documents using that anymore, at least I wasn't able to get that feature within them working

Since then I don't trust PDF anymore.

Imagine if something doesn't even display due to a missing feature... And in worst case it's something safety-related

I wish the standard made something like "This PDF document looks to make use of features not available on your system or PDF viewer" mandatory. Warning put in front and at the parts where the unsupported feature is included if possible.

Besides that most features are within Adobe's own proprietary PDF viewing software anyways (no surprise... even some of the 1.7 PDF features are or were proprietary not so long ago: <https://en.wikipedia.org/wiki/PDF#History>) and that is not even available for all major platforms even. I wish there was a competing format.

PDF and TIFF are kitchen sinks indeed

They can do everything and keep evolving, which is nice because new stuff like jxl can be added to them as a payload codec, but also not nice because you can never make a PDF or TIFF viewer that is complete and will remain complete

no but we could try with a usual one like 96dpi

Evolving file formats are nice but they should have really thought about what to do when viewers don't support a new feature

Or what to do when a feature is dropped...

Tho if there was something like a libpdf that might not have been necessary?

With some very rough maths and some dodgy memory, about 1 million KM squared

Max dimension in pixels is 2^40 in level 10 or 2^61 if you go outside the defined levels

If it's a square megapixel (the level 10 limit), and you have 100 pixels per centimeter then that's 10 km by 10 km or 100 km²

If it's the bigger limit (max size that can be signalled) then it would be about 2 million times more, so 200 km²

(you have to signal a height of 1 gigapixel for this, and a width computed according to aspect ratio 2:1)

So you would need two and a half jxl files to cover the entire earth

If it's 96 dpi instead of 100 pixels per cm, then one jxl file is more than enough to cover the earth (though not by that much)

wow

so like

let's make some jxl images of the earth?

OpenStreetMap, except it is a single JPEG XL image. 😄 Sounds like something DZgas would approve.

96 dpi is less than half the pixel density of 100 pixels per cm?

PDF/A is sort of this. Its a subset of PDF that is intended for archival use.

Okay, from wikipedia »All of the information necessary for displaying the document in the same manner is embedded in the file.«. That does sound pretty good

However, audio and video are forbidden

Well, guess it would be more complex and of course that stuff is not printable.

Accessiblity is another point for that I guess. Not impossible to describe audio and video but it's harder (with the exception of voice recordings which could be transcribed 1:1)

And then suggest that the colors are 3 shades off ;P

Mentioning gigapixel reminded me of these <https://www.earthcam.net/projects/empirestatebuilding/gigapixelpanorama/2021/> I remember a few years ago there was a lot of talk about a Chinese or Japanese based one where you had enough resolution to see people's faces from a mile away, nice to see they're still being made

Since this server is level 2, why not using a banner? <:megapog:816773962884972565>

no jxl support

Won't it lose its level 2 boost soon anyways?

idk 🤷‍♂️

I have a version of a copyright font that works well as iosevka 24.0.1.4

But I violated copyright

I was contacted on Discord

Sorry for the late reply (I usually search for libvips instead of VIPS in Discord, so this went unnoticed). The `depth` parameter of `icc_transform()` always defaults to `8` and isn't set automatically depending on the interpretation. To query the interpretation you could do: ```console $ vipsheader -f interpretation 16bitAlpha.png ((VipsInterpretation) VIPS_INTERPRETATION_RGB16) $ vipsheader -f interpretation 8bitAlpha.png ((VipsInterpretation) VIPS_INTERPRETATION_sRGB) ``` And set `depth` to `16` for `VIPS_INTERPRETATION_RGB16` or `VIPS_INTERPRETATION_GREY16`. /cc <@310374889540550660> perhaps we should set `depth` to `16` automatically for `VIPS_INTERPRETATION_{RGB,GREY}16`? This would make the following code unnecessary downstream: <https://github.com/weserv/images/blob/5.x/src/api/processors/thumbnail.cpp#L514-L515> <https://github.com/lovell/sharp/blob/main/src/pipeline.cc#L320>

Thanks so much for responding! And yeah, I naively thought I don’t need to check image characteristics and route different images through different code. I do think your suggestion would be a great enhancement!

> perhaps we should set depth to 16 automatically for VIPS_INTERPRETATION_{RGB,GREY}16? good idea! https://github.com/libvips/libvips/pull/3558

https://twitter.com/EP_AnimalLovers/status/1676793538196168705

AI covers seem to be the latest trend

https://youtu.be/6SOtyUdZiDE

<@288069412857315328> <@238552565619359744> in case ya wanna continue formulating the national content creation budget

Reading a bit of the conversation on <#794206087879852106>, I feel like I never really understood why people are so bothered by ads. Like, they are a bit annoying, but they help fund the platform and the people authoring content for it. Is it really too much to ask for people to take a few seconds every several minutes for that? It literally takes money away from the companies advertising and puts it almost directly in the hands of the people who create the media we care for. The only downside I can see is that YouTube itself gets a portion, but I don’t think that’s strictly bad. There are many people who work on things for the sake of charity, but it shouldn’t be a requirement. If a company or group wants to be rewarded for providing a platform for people to express themselves, I don’t think there is anything wrong with that, specially when those people are also being rewarded for helping the platform by providing media for it.

yes

I was never really bother by ads either however I started using a ad blocker because my browser and slow internet would start crying and dying from ads so I was sorta forced to use a ad blocker to making using the web not a struggle

on some smaller sites that only have like one ad or so I turn the blocker off though

I wouldn't mind ads if they weren't dumb and manipulative.

I don't want it taking my time. and it's too easy to install an ad blocker. So when it becomes not easy, I'm willing to pay to remove them

I don't like getting interrupted for something completely irrelevant to me

I block ads only if there are video ads on websites. They are very annoying.

For YouTube, I pay for youtube premium.

or popup ads

i use revanced ngl

I know. But I feel slightly guilty when using those apps.

I do not. I won't pay for premium until google changes their business model.

don't forget that this yt is owned by google

what do you want their business model to be

I want them to focus on and respect users, creators, and privacy. I want them to actually regulate the shit for kids rather than try to force popular channels to target a broader audience.

maybe if you give them money they'll do that

that's not a business model

and if you really cared about that stuff you wouldnt be using discord

:)

Claim: I do not care about respect for users and content creators because I use discord. Counterargument: none, the claim is flawed

First off, some content creators use discord almost exclusively to communicate with their fans.

Second, discord is not a content-creation and streaming platform. It serves an entirely different purpose.

i meant the users and privacy part but okay

Third, even if using discord was the worst possible action, discord provides a better platform than others.

How do you mean?

discord is arguably the worst when it comes to this

like they actually log everytime you hover over a button

and log your actual location (not ip)

that's... laughably stupid

when it comes to creators, youtube is probably the best

we live in a society

I wish matrix wasn't cringe

oh that's another option

self-host

matrix is cringe

fediverse/mastodon is cringe

what else

it's not the federation that makes them cringe

it's the implementation

If you dislike Matrix then XMPP, Chat over IMAP or IRC might be options

they also track the amount of time you spend in apps regardless of if you have this setting on (the only thing this info is used for)

There is also "discordo" TUI client for Discord

Which is usable at least

But there have been better predecessors imo that got shut down

Just because a user is giving up some privacy to company one with some connections to a chinese mega cooperation doesn't mean he needs to give up privacy to company two which is a american mega cooperation.

yes you do

google: amazon: china: tiktok: "meta":

yeah i submit to all of them

so you've finally revealed your true colors

smh

😔

When AdBlockPlus failed to block ads on Twitch a few years ago, I stopped visiting Twitch. Ads on local sites from my city can be actually helpful, but I still don't click on them: clicking on ads is an easy way to get malware. I instead look it up myself.

yeah, ads would be ok if they weren't literally computer aids

It's not even a new repository, it's a few years old.

usually the pay from ads is per click. If you never click on ads, you may as well have an ad blocker. And you know, you can charge for stuff, ads aren't the only option to make money. Unfortunately most platforms refuse to provide a way to pay a reasonable amount.

frankly, the entire online advertising industry must be destroyed

completely agree

Oh wait, I was wrong. Looks like search in forks is now possible! It's index-on-demand. You just have to try to search, then wait a few hours until indexing has ended.

it's subversive, misleading, a security nightmare, a psychological disaster, often the products are useless or already purchased, or the ads are just so, so bad and toxicly self-aware

another thing: youtube ads that pay "per view" are usually the ones you can skip after 5 seconds. skipping also negates any pay.

yep. blocking ads is basic self care tbh.

Trying to "help" creator by deliberately not skipping and clicking on ads is detrimental. Creator could be banned for fake views/clicks.

Exactly! Watching an ad feels like attending a work training. It's a segment of time that gives you nothing and actually might cause you to lose money on something you don't need. Or it just wastes time telling you about something you'd never buy (again).

even without that, keep in mind, advertisers like to pay based on how much actual busienss they get from ads. Increased views or clicks or whatever else makes the creator get more money, without an increase in people buying the thing advertised, makes advertisers want to pay less, negating any benefit of a person choosing to look at ads or click them.

And don't even get me started on the problem with tiered ad serving.

"But ads on your smart-TV make it cheaper!" argument simply does not work. If you let them, manufacturer would put ads AND make it more expensive AND add some bs subscription. Not to mention this argument implies a choice, which many manufacturers take away from you.

Better products become harder to find because worse products have more adspace.

Ah yes, the American dream: get advertised to in your comfort space.

targeted ads are a thing to make them not feel pointless

the issue in current day is that people are too cheap

youtube premium is an attempt to solve all of that

plus a premium view is huge compared to a regular view

but people feel entitled when they lived a while without paying for their entertainment

these concepts existed before the internet

going to the theatres to watch movies, free newspaper vs a subscription magazine

buying books

A creator on youtube makes what, 0.1 cents per view on average? I think youtube's cut is 50%, so say they get 0.2 cents per view on average. If I could pay 1 cent per view, that would be 5x more. Unfortunately, there is no such option, just a fixed subscription no matter how much or how little you watch.

(I do actually pay for youtube premium, but the value proposition isn't great)

my experience is the opposite. I have never, *ever*, seen an ad from a platform that tries to do targeting based on the information they collect that wasn't pointless.

I have on the other hand seen static ads that actually advertised stuff I might be interested in.

Most optimal ads are topical non-network ads that's ordered by site users themselves and self-hosted by site itself without external javascript. So furry ads on furry site, brony ads on brony site, web-novel ads on web-novel site, etc. Bonus points for users being sure that it's not fishing/malware due to site admins directly controlling ads.

idk it just feels creepy at that point

I have them disabled on every platform. They really get annoying. You look at 1 thing for 5 seconds and end up with 100 ads advertising that 1 thing for the next 3 weeks.

or you buy something on amazon and then all your ads are that exact thing LOL

like, what?

i already bought it just let me enjoy my memes 😭

I really don't get why this isn't the default. Or even just ad networks doing the contextual ads based on the page the ad is in.

https://bellard.org/nncp/

this is cool

This is why a lot of youtubers have switched to sites like patreon for funding

cause people actually will pay for content they like

someday i wont be lote

https://i.4cdn.org/g/1688959661219165.png

5 dollars a month for a couple of services would make those services orders of magnitude more profitable than they are today, so doubt

except for netflix maybe

without ad, federated services would have developped more (like IRC at the time) which is for me a win win situation.... communication and medias available without selling your soul to the devil ! 200x-ish internet with edonkey and IRC 😄

indeed. a lot of what I do online is already both free and ad-free.

Off topic: <@532010383041363969> - does zopfli test static Huffman tables for a `png`, or just dynamic?

Interesting, always found it hard to notice the difference between scaling methods, didn't realise there's that much of a gap (Even if it's just VMAF again) https://youtu.be/R6Ly0GYlXa4?t=317

just use nnedi3

had to check chrome image scaling and it is still bad

only dynamic

Thank you! That helps with a discussion on Oxipng here: https://github.com/shssoichiro/oxipng/issues/420

Hi <@532010383041363969> - there is a section of code in `zopfli` that seems to check static blocks?... https://github.com/zopfli-rs/zopfli/blob/109d21174e637b8c0a2167d67bc381c739e6ecab/src/deflate.rs#L1082-L1129

https://www.youtube.com/live/q2ueCg9bvvQ?feature=share

liftoff in about 20 mins

Lunar Exploration Mission

https://youtu.be/8dgeOPGx6YI?t=7m44s this would be easier to follow with less compression artifacts

reminds me of this video I once uploaded to youtube: https://youtu.be/GViAq_IWOg4

I had to create a new video, with smaller sandpiles and 2x NN upsampling to get something that would somewhat survive youtube compression: https://www.youtube.com/watch?v=WmP189_bxbc

https://discord.com/channels/794206087879852103/1020056709831594005/1130291480766980116 to continue this: it's been 7 months since I've got a new job, and I was pretty busy working on various things

"various things" here include: working on a large Rust codebase, creating a new Python binding for that Rust codebase (and both Rust and Python involved a heavy dose of async), debugging various problems over the course (I've contributed to PyO3 and cargo-zigbuild), documenting the whole things, revamping the whole CI pipeline to allow nightly builds of that Python binding using Tekton

to be fair I really enjoy the whole endeavor, but also means that I need to concentrate a lot and I'm left with no spare energy at the weekend

Makes sense!

oh, thanks for checking

We're about to eat this cake

you mean exterminate it?

😋

The camera was edible too

nice

A meeting about this I presume https://jpeg.org/jpegtrust/index.html Ah, 100th JPEG meeting after opening the image properly, conrats!

The meeting takes a full week, many things are discussed

For jxl we discussed the 2nd edition of 18181-3

Nice touch with the camera strap too, surprised it held together so well

> with a tendency towards significance for Group 3 (P=0.147)

https://tenor.com/view/long-jump-brittney-reese-team-usa-nbc-olympics-running-gif-22949165

https://youtu.be/41nN9i7y2_0?t=2m14s technically impressive for a GBC game but a bit nauseating nonetheless

----------------- > 🥱 -----------------

my own pure ffmpeg noise-color filter for VHS like destruction ```-vf "hqdn3d=1:50:20:100,noise=alls=50:allf=t+u+p,eq=saturation=1.4:contrast=0.9,colorbalance=bs=.2,histeq=0.02:0.25:none"```

And just as this got merged too https://github.com/libjxl/libjxl/pull/2675

Seems they update the canvas on the client by sending a 1000x1000 png with new pixels every frame and transparency elsewhere... Safe to say JXL cuts that in half even on effort 1 (Lossless naturally)

Another time?

Interesting idea, implementation could've been better though https://youtu.be/CuNgLZiWkm8

anything new or just clarification?

Do know why autocorrect @'d a random person there

Still in early draft stage, it's just about conformance testing so not super exciting, but we'll probably restructure things a bit

> Perhaps this re-brand is just Elon putting his weight in the X11 vs. Wayland debate? lol

lol

ngl the X.org logo is much better

gl twitter is good

now he just needs to start companies called S, E, and Y

Why X.Org when it it is the libx11 logo that reminds people? On Twitter "X Window System" was in the trends in several countries even and most people used the correct logo, did not mention X.Org and did not use the X.Org logo.

don’t you mean just X11, without “lib”? libx11 implies an implementation rather than the protocol, and the current dominant implementation is, as it happens, X.Org

https://gitlab.freedesktop.org/xorg/lib/libx11

oh, never mind, somehow I was under the impression that the “naked” logo was for X11 itself

Probably both, I just wanted to include a fun fact.

repo for libxtrans has trans flag as an icon btw.

https://youtu.be/KStJfpHsImE?t=141 Looks good

Hi... asking a favour of anyone running MS Windows - with a *non-English* locale! Please could you run `cmd` and then copy&paste the result when you run `echo %time%`... I'm trying to settle whether this format changes depending on your regional settings! Thanks in advance 🤓

You should be able to change your computers locale and then chance it back

```console 0:08:20,20 ```

⬆️

That's great! Thank you. What locale do you have?

fr-ch

wait, I might have customized a few parameters of it

I have it set to use “,” as the decimal separator, which I don’t think it does by default

(fr-fr does)

That's no problem.

And safe to assume the time is HH:mm:ss.hh (24 hour with no leading zeros, all others have leading zero)?

I suppose so, but do you have an English Windows to compare to? then we can be more sure of it

I want to gather as large a sample as possible 😄

There should be a way to get the ISO UTC time

Which is internationally standardised

My clock isn't set to the same time... but looks like it. You can run the command a couple of times to see how the digits change.

Thanks agin for the help!

Hoping to get some results outside of latin alphabet regions... 🔍

``` 8:50:12.83 ```

(ko-kr)

That's great. Thank you! And safe to assume the time is HH:mm:ss.hh (24 hour with no leading zeros, all others have leading zero)?

yeah, it now prints something like `17:...`

Twitter renaming to X must be the worst branding decision ever

It's a rare thing that a brand becomes a verb (to tweet) and noun (a tweet) that is in common use. Throwing that away to replace it with something as generic as "X" is insane.

I'm considering to buy a new phone. I usually go for mid-range (not €800+ flagship models, but also not cheap but kind of crappy < €200 models). Does anyone have advice?

OnePlus Nord 2T 256GB is €400, seems ok on paper?

everything seems ok on paper. but literally everything other than samsung, google, and apple lose support so quickly

rename jpeg xl to jpeg twitterl

I'm also considering fairphone and nothing phone, for something a little less generic. More expensive though.

I gave up on alternatives and will only go samsung flagship and free google pixel

time and peace of mind is very valuable

must be nice to be in the apple ecosystem

a friend with a fairphone 3 got a notification that the android 11 -> android 13 is ready but that he’ll lose fingerprint functionality with some apps

I actually switched to an iPhone ~2.5 years ago for that reason

my Xperia XZ2 Compact stopped receiving security updates, less than two years after I bought it

so I thought “all right, enough of this”

I use a mid range Samsung, which is fine and has decent support (3 major android versions and ~5 years security.) Apple is definitely the easiest to recommend, because they don't produce any bad phones. Native JXL support is also a plus. You can probably do better with a lot of shopping around on the android side though.

iPhone is very expensive imo. An iPhone 14 Pro Max at ~1400 EUR has about the same hardware specs as a ~400 EUR Android phone (though I guess the walled garden / tight integration software approach does allow it to get a bit more out of the hardware...)

thing is you can also resell the 14 pro max for 80% the cost in 5 years

I liked my SE (except the battery life)

speaks more to how long they last so it's like 1400 eur over how many years vs alternative cheap phone over how many years

I have a lot of custom apps I sideload now but when I had an iphone 4, I had no complaints

Mine is 6 years old, battery is sad and stopped getting firmware updates after the first year, then suddenly got one about a year ago... That removed half the functionality of the camera and added 'AI Search' along with half a dozen more bloatware apps. It does seem to run a bit better lately though

I dislike the walled garden enough that I can't use an iOS device anymore unless I jailbreak

but at that point might as well use an android

I currently have a Samsung A51 5G and I have no real complaints

Buy nothing phone 2?

1000EUR surcharge for jxl support <:NotLikeThis:805132742819053610>

> An iPhone 14 Pro Max at ~1400 EUR has about the same hardware specs as a ~400 EUR Android phone I really really doubt that this is true. Maybe equivalent on *some* parts of their respective specs. Would you happen to have any particular €400 phone in mind? I use an about €500 phone (sony xperia 10 iv) and it is lacking in many ways, while being very good in others (3 different zoom lenses is unusual at this price, but the picture quality is not great. the processor is slow enough to be noticeable in many apps, but the battery lasts forever)

All in all it depends what you want out of your phone. For my use case I just want it to be fine.

If you just want straight recommendations, the Samsung A54 is probably fine, or the A34 if you want a smaller phone.

If you got for more niche manufacturers you can probably do better on the specs, but probably also have to worry more about random issues.

I don't think I ever want to use Samsung again

My significant other uses Samsung, I don't want their Android customizations tbh

i want just like, a normal os

why do mobile os have to be so weird

motorola is good, it is almost plain aosp

that's kinda the problem- if i want to report a bug on aosp, i still have to go through google plus, they fucked up the SDKs and NDKs so bad it's basically impossible to do any app dev without their build system

I don't understand

imagine you want to make an app for linux or windows now imagine doing the same thing on android

Okay. I do not have problem using android SDK. so I don't understand the problem.

o

it just seems like a lot to learn to do what is essentially webdev

they created a monster of an ecosystem that forces devs to work around it instead of the other way around

idk maybe im just impatient

you try getting into something and the very first thing you try suddenly turns red with errors

xD i sound wimpy

the Fairphone will get support forever, & the Nothing Phone has a great software experience from what I've seen. Are Pixels available in your region?

In the US at least, Pixels from just one or two generations ago are *ridiculously* cheap for the product offering

yep

why so much resistance to the international phonetic alphabet

most nonlinguists don't know it

is probably the reason

idk looks fine to me

Pixels are available here, but not as cheap

384 EUR for a Pixel 6a

509 EUR for a Pixel 7a, 869 EUR for a Pixel 7 Pro

How much for a used/refurb Pixel 6?

Actually can find one for 339 EUR too

307 for used

Still using my xz1 compact. I care more about my e-wastes than about a potential security threat... Mainly because my phone is so full of crap already, a virus couldn't find a single byte free to install itself 😅

That's pretty good, esp considering buying used is more eco-friendly

I've found that the Pixel 6 tends to be cheaper than the 6a due to being out longer or something

7 was free for a while in NA while on most providers' plans. those are crazy prices

it taught me that samsung flavor of android is pretty necessary

I don't like the Samsung UI. Its icons look childish.

I don't like that it defaults to having the back button on the right, but at least you can change that

I went for a Motorola Edge 40, which was 437 EUR which is cheaper than a Pixel 7 while it looks about the same or even slightly nicer to me

I've been getting into straightedge and compass construction, but instead of normal stuff I'm doing stuff like this. Equally spaced horocycles on the hyperbolic plane, projected to a Euclidean surface with the Poincaré disk model

Pixel 6 has as good cameras as 6 Pro, just no telephoto. 6a has a shit main camera.

out of curiosity are these lines on the poincare disc? I'm not super familiar with hyperbolic geometry

Lines would be circles (or arcs, rather) that intersect the edge of the disk at right angles.

ah, what are these then? they clearly go off to infinity

so they're not "circles"

I suppose they kinda seem like elipses that have one vertex at infinity, so maybe parabolas?

parabola being defined as having a focus and a directrix

and the locus of all points that are equidistant from the focus and the directrix

With the way the Poincaré model works, hyperbolic circles look exactly like Euclidean circles, albeit with a different center point. Since the edge of the disk is infinitely far away, all finite objects do not intersect it. A horocycle is what you get when you take the limit of a circle as the center goes to infinity. They have infinite volume but curve just slightly less than a circle would. The theoretical center point at infinity is called an "ideal point". In the Poincaré model, a horocycle is a Euclidean circle that touches the edge of the disk at a single point.

well in euclidean space, if you fix a point on a circle at the origin and take the limit as the center goes to infinity, then you end up with a line

yeah

a line that goes through the center, 90 degrees off of the direction of the limit

that's why I was wondering if these were lines

mm

so you're saying that in euclidean space, the limit of a circle is indeed a line as I described

but in poincare space, the limit of a circle looks like a euclidean circle that is tangent to the edge of the disc

yeah

but in poincare space, it's not a circle

it's got no analogue in euclidean space

yep

and the reason it can "curve less than a circle but more than zero" is because of the nature of hyperbolic space

that's cool

in euclidean space, all curvature can be described in terms of a circle's curvature

i.e. local curvature for a C-2 curve can be described as 1/r where r is the radius of the circle that agrees up to the 2nd derivative

but it sounds like you're saying that's not the case in hyperbolic space

there are also hypercycles, which are curves which are equidistant to a line. In the Poincaré model, they look like arcs that intersect the edge at other angles than 90°, at the same points as line they follow. There curve less than a horocycle, and if you take the limit a hypercycle that passes thru a point but move its line towards infinity, you *also* get a horocycle. So horocycles can be seen as the exact mid-point between circles and hypercycles.

huh

in euclidean space a hypercycle is just a parallel line

which makes sense, since parallel lines in euclidean space remain the same distance apart everywhere

but parallel lines in hyperbolic space get infinitely far apart in the limit

yeah. I'm not sure on the formulas, but in hyperbolic geometry, the circumference of a circle grows exponentially with its radius, and if you were to walk around a large enough circle, it would feel as though you had walked around it many times before you finish your first revolution.

Someone on math stackexchange says that in Euclidean terms, a hyperbolic circle has curvature 2/tanh(R)

There is another way to measure curvature which works on any Riemannian manifolds, called geodesic curvature

Unfortunately I do not understand the wiki article <https://en.wikipedia.org/wiki/Geodesic_curvature>

Fortunately, the stackexchange answer says that if the point whose curvature you wish to measure is placed at the center of the Poincaré disk, then the geodesic curvature is simply half the Euclidean curvature.

A line (geodesic) of course has a geodesic curvature of 0. A hypercycle has a curvature > 0 and < 1. A horocycle has a curvature = 1. And a circle has a curvature > 1.

interesting

curvature in euclidean space is `1/r`

so it can be any positive real number

Anyone else going to the Open Source Summit Europe in September? https://events.linuxfoundation.org/open-source-summit-europe/

<@853026420792360980> I wonder if you might appreciate this... it turns out that the infinite hyperbolic plane can be conformally mapped to any solid shape, and someone did the math to create this: https://www.youtube.com/watch?v=P9EZy8ZzlVU

I hate this

in the sense that I love this

😄

why is `google-cloud-cli` such a huge package? it's the biggest debian package I have

/usr/lib/google-cloud-sdk is taking 923 MB

this seems huge overkill for, you know, some CLI tooling for _something something cloud_

I see a 81,510,105 byte file called `gcloud.json`