#Kagi Translate is so much better than Google Translate. So much more detail and explanation of what the text might have meant depending on the context.
Remote
Jeffrey Yasskin
@jyasskin@hachyderm.io
mastodon 4.7.3I work on #WebStandards for Chrome, and I'm organizing Google with the #AlphabetWorkersUnion. I'm an appointed member of the #W3CTAG. I miss programming, but now I mostly talk to people. #Urbanist, pedestrian, #YIMBY. White, but trying to check my privilege.
Opinions here are not anyone's but mine.
614 Followers
252 Following
48 Posts
Joined November 06, 2022
Pronouns:
he/him
Github:
Location:
Portland, OR, US
RE: https://timeloop.cafe/@sus/116932130441075196
It's amazing that a person who works at Palantir and loves AI would be so offended by the enshittification of physical goods, but he seems to be passionate about something useful in addition to the terrible things:
Presenter at BlinkOn today: The first issue asking for this feature was filed in 1998, before I was born. The current issue was filed in 2018, before I was in college.
Me: 👴💀
Replying to
@jaffathecake@mastodon.social @slightlyoff@toot.cafe Speaking as an individual TAG member, we should be laughing at the notion that a developer could be seen as agreeing to a ToS by writing a line of JavaScript. It's like if I put in my profile that "if you write a post I wind up seeing, you owe me $10". Big companies that believe absurd things can still use the legal system to cause problems, but we shouldn't lose track of the fact that it's absurd.
I wonder how "You Only Live Once" came to mean "throw caution to the winds" instead of "make sure not to mess up that one life."
I hate AI-generated slides. Most #BlinkOn presenters have the sense not to use them, but the others generate extremely busy slides, in which I can't tell which text and images are actually relevant to the presentation, vs which are just fluff.
Replying to
@sphcow@mas.to I'd quibble about the IETF levels, at least compared to the W3C. My impression is that IETF Internet Standards are more mature than any formal level that the W3C produces, and that IETF Proposed Standards match W3C Recommendations. You're the expert on the IETF side, of course: have I missed something?
It's also easy to get confused between the various kinds of IETF Internet Drafts. Since an individual can publish one unilaterally, some have less maturity than a W3C Community Group document (although an equivalent IPR grant). Those are easy to confuse with the WG-adopted I-Ds, which, as you say, are equivalent to W3C WDs.
Replying to
@thibaultmol @ireneista @erik @vikxin @NanoRaptor And they have a "report slop" button to crowdsource detection, with the crowd's quality increased by the fact that the whole crowd is paying them. https://blog.kagi.com/slopstop
(I work at Google, and a too-strong market position is as bad for the workers as a too-weak one.)
Replying to
@freddy Did you learn any lessons that would apply at the specification level? Totally reasonable if these were all implementation bugs, but I'd love to learn about mistakes we're making at other levels too.
https://www.portland.gov/transportation/walking-biking-transit-safety/bike-bus is a neat initiative, but illustrates the disconnect between schools and working. My kids' school accepts drop-offs between 8:15 and 8:30. I think the rest of the PPS schools are similar. The city bike bus arrives downtown at 8:15, so nobody responsible for kid drop-off can participate.
#pdx #school #biking
Trying out a new TODO list system. Currently in the ADHD phase of "this is new, so I'm using it" with an unknown amount of time until either "this actually works" or "this is old and boring so I'm not using it."
Replying to
@jaffathecake@mastodon.social @slightlyoff@toot.cafe https://groups.google.com/a/chromium.org/g/blink-dev/c/iR6R7-nQeHI/m/9Sb7o12jAgAJ lists some more sources for developer feedback, but it looks mostly private, so it's all about whether you trust the team to grade their own homework, which we generally shouldn't. Even if it were public, I don't think I'd have the expertise to tell whether it was net positive, given the selection bias introduced by hosting a hackathon.
Replying to
@mnot 2 thoughts:
-
If an origin contains many logical websites, as in the old example.edu/~user/, well-known becomes inappropriate, or at least needs a way to dispatch to the logical units' feed lists. App Manifests would be a way to deal with this.
-
Re the localization discussion downthread, App Manifests made the "just content negotiate" argument for a long time, but have recently added inline translations. It'd be worth checking whether the same reasons apply.
The W3C @tag@w3c.social is going to Hong Kong in 2 weeks for a face-to-face meeting, and as part of that, a local developer group is hosting us for a developer meetup: https://www.meetup.com/codeaholics/events/310842831/
Wednesday, September 17, 7:00–9:00PM on the 7th Floor of the GoodNotes Office at 8 Observatory Road, Hong Kong
If you'll be in Hong Kong then, I hope to see you and answer your questions! Please boost this and let more people know about it.
Draft points opposing https://www.congress.gov/bill/119th-congress/house-bill/8250/text, the U.S. age verification bill: (Aside from the first two, I've largely copied these from other people.)
* Responsible parenting requires us to have a discussion with our kids about what kinds of content and experiences are appropriate to them. Not all 8-year-olds are ready for the same content, but this bill requires a one-size-fits-all approach.
* Even if the bill allowed parents to give our kids a different age for each app, some 8yos might be ready for the average 10yo's reading level but only the average 6yo's level of scariness. Age is just the wrong value to use for this kind of configuration.
* In order for a parent to "decide" under this bill, we're going to need to lie to our kids' devices about their ages, which sets a bad example to our kids.
* Abusive parents should not be allowed to decide for their kids.
* As an adult, I shouldn't need to have an ID or to submit data to an online service to be able to use a computing device.
* As an adult, I should be able to modify the software running on my own devices without being required to connect it to some online service.
* Kids should also be able to write and install their own operating systems.
* Once operating systems are set up to verify one aspect of their users' identities, a new administration can much more easily come in and demand they restrict use based on another aspect.
What else am I missing?
Replying to
@ireneista My impression, and I think @npdoty 's, is that the California law makes OSes *ask* their owner for the user's age, in order to pass it on to apps. It doesn't make them verify that age, so it facilitates parents working with websites to give their kid an appropriate experience. If a kid is more mature than their age implies, or websites are hiding age-appropriate information, the device owner can say a different age. Seems in line with https://www.w3.org/TR/privacy-principles/#guardians Much better than the jurisdictions that are requiring sites to verify with private-info uploads.
I could be missing something, of course.
2 #Web problems which seem like they might share a solution:
1) You link to a page, but there's a chance that it's gone offline or has a too-steep paywall, so you want to also link to an archive site like web.archive.org or archive.ph. It's possible to put those links adjacent in the HTML, but that stops the UA from doing something intelligent if it receives a 404 or the user otherwise expresses that the main link didn't work.
2) https://www.tbray.org/ongoing/When/202x/2025/04/16/Decentralized-Schemes and https://github.com/mastodon/mastodon/issues/19679#issuecomment-1301180085: you're supposed to be able to link to a scheme: url to get to the user's choice of application to handle that scheme, but this leads the user to a dead-end if they have _no_ application for the scheme.
An obvious-solution-that's-probably-wrong: Add a fallbackhref="space separated urls" attribute to link tags, and have browsers automatically load them in sequence if earlier ones "fail", for a definition of "fail" that includes at least 4xx and 5xx HTTP errors and missing protocol handlers.
Edit: Initial very-drafty explainer: https://github.com/jyasskin/handling-unreliable-links
Replying to
@slightlyoff@toot.cafe @jaffathecake@mastodon.social Please do let me know if anyone at Google actually called this a "web standard". It's not shipped widely, so it's not a "standard", and we've been trying to stamp out misuse of the term, but people do still make the mistake.
Replying to
@jaffathecake@mastodon.social @slightlyoff@toot.cafe Those aren't limited to developers, right? Was the reaction to the Prompt API more or less strong than the reaction to Firefox's AI features? I personally share the "hate everything to do with AI" sense, but I think the API owners would focus on developers actually building sites over random people with opinions on the internet.
If you look at https://api.github.com/repos/webmachinelearning/prompt-api/issues/74/reactions all but one -1 is April 30 or later, a month after the I2S. That's an internet mob, not developer sentiment.
Is it "WPT tests" or "WPTs"?
@fugueish@wandering.shop And cited from there, https://lpeproject.org/blog/what-would-a-russell-vought-of-the-left-look-like/ is good. "two entrenched sites of hierarchical power: juristocracy and the modern managerial presidency. A reconstructionist constitutional politics would not merely lament these structures; it would seek to destroy or transform them. ... We must work to abolish presidentialism in the hearts and minds of the American people."
Replying to
Replying to
@jaffathecake @gundersen That idea reminds me of @PatMeenan's compression dictionaries... I wonder if they could share infrastructure...
Replying to
@npdoty @ireneista Reading https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB1043 I'm worried about 3 things:
1) '"Account holder” means an individual who is at least 18 years of age.', and the OS has to ask "the" account holder, but there's nothing about how the OS figures out its administrator's age. Presumably it'll be ok to assume that the person setting up a device bought the device, and only 18yos can buy things, even though that's strictly false?
2) The law talks about "the" user, so what about devices with multiple users? (probably minor?)
3) The law assumes that the OS will give apps a granular version of the single age specified for the user. But a good parental control app ought to let parents say "my kid is Xyo for this particular app". Is that allowed?
Replying to
@hvdsomp @pointlessone It took a while to find, but https://github.com/iipc/robustlinks seems to be the implementation described there.
Replying to
@tim@union.place @jefftk@mastodon.mit.edu I do it via a donor-advised fund, as Jeff suggests, and it's pretty straightforward that way. The hardest part is when the organizations don't put their tax id on their site, and they're incorporated as a different name or taking donations through some fiscal sponsor.
Replying to
@vaurora@mstdn.social Silly response: Is "ghost" the right word for these? They weren't alive and then died-but-stuck-around. Instead, they're ... spirits that were forced into existence without ever being born. Is there a mythology that gave a name to that kind of thing?
I have an LDS friend who mentioned that their theology distinguishes pre-mortal spirits from kinda-demons, and you can tell the difference by whether they agree to shake your hand and then fail. Since these projects do try and fail to interact with you, maybe they're demons.
Replying to
@hdv@front-end.social I think we do need to get rough estimates of the impact. If it's a 0.1% increase in total CO2 emissions, we should focus our climate efforts elsewhere.
Replying to
@torgo@mastodon.social Firefox and Safari seemed fine with this approach. ("smaug: this looks like a reasonable proposal so far" in https://lists.w3.org/Archives/Public/public-editing-tf/2025Mar/0000.html) They'll do another review once there's proposed spec text.
Replying to
@torgo@mastodon.social There's a discussion *in 45 minutes* in the Editing WG about a new feature that would accomplish the use cases that previously needed this permission-guarded any-time access to the clipboard contents. They really need any-time access to the _types_ on the clipboard, like iOS allows for apps without a warning, and we'll see how Firefox and Safari feel about adding that to the Web.
Replying to
@fastmailunited@union.place Hooray! I just started using @fastmail@mastodon.social, and having a union makes me more likely to recommend it.
Replying to
@hvdsomp That's still data-versiondate. The ability to provide dates in data-versionurl seems to be new in the latest version of https://robustlinks.org/spec/#data-versionurl What do you expect the next version of the JavaScript to do with it?
Replying to
@hvdsomp I definitely see the use for versiondate, maybe even in helping a user notice a hijacked site. How would a UA use the embedded memento dates in the versionurl value?
Replying to
@tim@union.place It probably depends on the DAF, but for Fidelity, "Nonprofits that have never enrolled in our EFT program receive grants through checks mailed to the addresses listed on your organization’s IRS Form 990." from https://www.fidelitycharitable.org/nonprofits/faqs.html I'll send you something when I get back to a laptop, to see if that works out.
Posted a description of why the ePrivacy Directive is broken, and why that part of the proposed "Digital Omnibus" is moving in the right direction, but doesn't actually fix the problem yet: https://jeffrey.yasskin.info/posts/2026/04/eprivacy/
RE: https://infosec.exchange/@sophieschmieg/116363970607065790
I love the Passover reference at the bottom of Scott's post.
Replying to
@cassidy Is the image important for readers to understand the post? If not, omit the distraction, and try to ensure the link preview doesn't have one either.
Replying to
@qgustavor@urusai.social @torgo@mastodon.social The other browsers now handle this by showing a little [paste] menu behind your mouse pointer when the site tries to programmatically paste. It's clunky, inaccessible, and should let users allow-list new keyboard shortcuts, but at least it enables the functionality.
I need to get better at going to bed on time if I'm going to keep scheduling 6am meetings with Europe. 🥱
@fugueish@wandering.shop Was going to shill Kagi, but their first result is CVE-2025-15435 (i.e. wrong year). The second result is correct. Quotes don't uprank that second result for some reason. It's still the first day, so maybe their crawls aren't as aggressive?
Replying to
@jaffathecake@mastodon.social @slightlyoff@toot.cafe It did represent developers as "strongly positive" in the I2S at https://groups.google.com/a/chromium.org/g/blink-dev/c/iR6R7-nQeHI/m/gN4iEGEdAQAJ and "positive" in https://github.com/webmachinelearning/prompt-api/blob/main/README.md#stakeholder-feedback
I always have trouble figuring out what developers think in aggregate, so I don't know how to evaluate the disagreement between that assessment and your "mostly negative" one. Where did yours come from?
Replying to
@mandymichael@front-end.social If maybe-ignorant questions are ok, I have two:
1) How much does my website have to vary its fonts before the download size of a variable font pays for itself? e.g. if I just use bold and regular, I think I'm better off just using 2 non-variable fonts? Does it tend to switch over when I add italic, or do I need to use a bunch of weights before it's worth it?
2) If I want to use the font-variant-numeric options with a variable font, what are my options?
People use "disintermediation" for both removing intermediaries (correct) and adding them (incorrect). It's like how "literally" now also means "figuratively".
Huh, the Seattle Link train gets up to 1600ppm CO2 when it's moderately-full.
Replying to
@chrysn Interesting: `rel=alternate` isn't body-ok on its own (https://html.spec.whatwg.org/multipage/links.html#body-ok) but it can be an `itemprop` for an object declared on an ``. [Edit: er, oops, s must have exactly one of `rel` or `itemprop`.] I can imagine that some RDF property would express this, although because RDF edges aren't ordered, there would have to be an independent definition of what order to try the fallbacks. I suspect it'll wind up too verbose to be useful, but it makes sense to describe it as an alternative to consider. Want to send a PR to https://github.com/jyasskin/handling-unreliable-links
Replying to
@hvdsomp Thanks for the prior art. Here's a really early explainer for the idea: https://github.com/jyasskin/handling-unreliable-links