It doesn't help that most hospitals essentially objectify patients these days.
When you are a patient, you are not a person, you are an object of medical care. There is an algorithm, a schedule; the arrival of new doctors and nurses at each shift change. They meet and talk with each other about how their objects are doing, and then they'll check in on you with a formal hello, 30 seconds of conversation, and then they'll leave.
If a patient has signs of mental illness, the providers draw straws to see who has the unlucky job of: "talk to the patient, but gently remind them that they are an object on a medical assembly line and probably nothing is going to happen until the next doctor comes around".
I don't know if hospitals have always been this way, or if we just romanticized them in the old movies. But unless you have a specific problem that medical intervention can solve (e.g., a medicine or a procedure), you are effectively a nobody, and there is nobody there that cares about you - at least for the definition of care that matters to most humans.
Can’t imagine Google is making the process of obtaining source code easier on themselves though.
Android has always been more source-open than “open source”. The vast majority of community contributions that make it into the codebase are security fixes and small bug fixes.
Everything else is essentially all the work of Google and (to some extent) Samsung.
GrapheneOS is arguing that throwing away the metadata of however many commits and squashing them into a messy tarball is not the "preferred form of the work for making modifications", and that a manual process where you have to fill out a form in order to get a Google Drive link a week later is not "a medium customarily used for software interchange" in current times. Those are quotes from the GPLv2.
If distributing under 3(b) then it's legitimate to only supply source on request. Historically source has been distributed without revision control history or metadata and been considered acceptable (the source tarballs on gnu.org are snapshots, for instance) so I think the preferred form argument is also tricky. I agree that there's huge value in having the individual commits, but from a GPL perspective we had this argument when Red Hat started flattening all patches in the RHEL kernel source 15 years ago.
When the build system expects a certain metadata which is removed by the force pushes or tag removal, then that is arguably not the preferred form. Grapheneos notes this elsewhere in the thread. [1]
I've been wondering about the Red Hat model! Seems like so much of modern development involves git blame or whatever to make sense of how the code came to be, and sometimes rule out an "obvious" modification that actually turns out to be a bad idea now that you know the historical context.
GrapheneOS might have a better case though if it's not just about understanding the code but about how the Android build system expects that everything is in Git.
I don't know what "Red Hat model" you mean, but if you're referring to their code delivery: It's not a problem for their internal developers, because they can use git lol. You can theoretically go to the upstream projects individually and merge the code drop with their nearest branch, and THEN do git blame/diff to figure out the delta and history of some lines of code. There are probably a few upstream open-source projects out there with no public VCS, but those are rare.
It's an impossible stretch and also a bad idea. Insisting on that interpretation opens up other issues. Revision history may not be available in some cases for various reasons, such as if one pays someone for a large contribution. We could get into arguments about how granular commits need to be to comply with publication requirements. Digital achives containing source code files are certainly customary and easy to use, as opposed to reams of printouts. It would be just as customary and modern to publish on CD and mail the stuff out to anyone who demands the code. Continuous or timely delivery is not guaranteed by the license. One week of wait time is actually reasonable for a process like this, though I expect it might gradually get worse in line with their long-term objective of making the entire thing painful for outside developers.
Tarballs certainly aren’t in violation of the GPL, via gnu.org own source distribution methodology. It would also be a very interesting legal claim that git metadata is required, and likely something already hashed out when the government decided that metadata isn’t the same as the underlying data.
Lack of a specific time limit in GPL doesn't mean there isn't one based on what's reasonable. What they're doing it not reasonable.
> Android has always been more source-open than “open source”.
This is about Pixels rather than AOSP. Google decided Pixels would no longer be supported by AOSP which is why they stopped pushing the kernel drivers, userspace drivers and other Pixel related code to AOSP. They moved to publishing kernel driver code via Google Drive after filling out a Google Forms submission. They're handling those manually. It's a ridiculous system and comes across as them wanting to make it a hassle on purpose. Perhaps that isn't the case and someone simply needs to make the decision to simply push Git tags somewhere else. They could put it on GitHub if the goal is disassociating it from AOSP.
We accepted the archaic new system while they were responding to requests in a reasonable time but that ended. It's now very inconsistent and is regularly getting delayed for weeks or more.
> The vast majority of community contributions that make it into the codebase are security fixes and small bug fixes.
Open source does not imply anything about accepting contributions. SQLite barely takes any contributions and the same applies to many projects. There were a lot more code contributions to AOSP than you're describing prior to recent changes with Android 16. It was relatively easy to contribute to the lower level parts of it.
Google are absolutely within the letter of the GPL, pedantically so. But maybe not the sprit.
We didn't have git tags when GPL was written in 1989, and while we did have sccs and rcs, (and early versions of cvs) they just weren't that widely used, and generally not used for distribution.
Even when GPL 3.0 was written in 2005-2007, source tarballs were still the primary form of distribution, even though it was starting to become standard to additionally provide anonymous cvs, svn, or one of the brand new distributed systems like git.
But these days git is the primary form of distribution, and source tarballs are noting more an afterthought. Hell, even tags are a bit of an afterthought on many projects. It's basically become the norm to expect an healthy revision history for any open source code.
Based on it's stated goals of "freedom to modify the software you use", IMO if the GPL was written (or updated) today, it would most likely require the distribution of revision history and restrict how much that history can be squashed/rewritten.
> We didn't have git tags when GPL was written in 1989
The GPLv2 requires "a medium customarily used for software interchange", not "a medium customarily used in 1989 for software interchange". Customs are the customs of the time someone is releasing the work.
“a medium customarily used for software interchange“
Interesting choice of cropping for that quote.
If you had included the previous word, it would be blindly obvious that “on a medium“ is only talking about the transport layer, not the format of the data. So it would exclude distributing source code on tape, or even optical discs, as nobody uses those anymore. About the only medium used for source code distribution these days is “the internet”
You could potentially stretch this to excluding google drive, though google will argue that the medium is http, not google drive; But you can’t stretch this to requiring it be formatted as git.
And even if you did manage to successfully argue that, google would just ship it as one (tagged) commit per release. The history would still be missing.
> If you had included the previous word, it would be blindly obvious that “on a medium“ is only talking about the transport layer, not the format of the data.
I can see your point. But OTOH if providing the "preferred form of the work for making modifications" requires not squashing commits into a big mess to frustrate someone trying to make sense of the code (I'm sure the Google engineers making modifications prefer to look at individual commits!), and Google is using Git anyway to create those commits, then it seems to me like there's an argument to be made that the customary medium used to interchange a range of Git commits is Git.
Yeah, that argument is much better; Maybe google are violating the GPL.
What that part of the GPL does absolutely forbid is stripping out the comments from the source code, or only shipping generated source files (but not the code which generated it). And arguably, git history is much of the same thing as comments. Especially when we have IDEs that can be configured to give us easy access to blame to help us understand the code.
The fact that commit history is not inline with the source code (or even stored as human readable files) makes it a little harder to see/argue that git history could be considered to be part of the source code, and I do wonder where this argument stops? Should the contents of bug trackers and PRs be considered to be part of the source code? What about design documents?
GPL says we need to be provided with the preferred form for modification. Those are the Git repositories for Android source code.
For Pixel kernel drivers, Google is converting the source code from the preferred form of dozens of Git repositories to a massive monolithic tarball. The build system used for this code expects it to be a bunch of Git repositories and spews out errors without it. It does build the code but it isn't the build process they used to do it themselves which is non-compliance. We need to be provided with what is needed to build it in the same way they built it.
> Android has always been more source-open than “open source”. The vast majority of community contributions that make it into the codebase are security fixes and small bug fixes.
I believe this isn't actually about "Android" at all but rather Pixel. Android is still openly accessible on git. But the kernel sources for Pixel devices is now behind this big song & dance for some fucking inexcusable reason.
It's about the Pixel kernel drivers and build system. The source code for the base kernel tree itself is still part of AOSP. Everything related to Pixels is no longer being pushed to AOSP.
They stopped pushing tags for any of the Pixel kernel or userspace driver repositories to AOSP. They also stopped pushing AOSP releases specific to Pixels which is why AOSP now only gets yearly releases, QPR2 releases and security backports to both of those. Other OEMs are meant to use the yearly and QPR2 releases along with the security backports to those so that's all they push. The monthly and QPR1/QPR3 releases aren't used by their OEM partners so they stopped pushing them to AOSP. Those no longer being pushed is because of them deciding AOSP doesn't support Pixels anymore.
Open source doesn't imply open to contributions. And you could imagine source available software that's not open source but takes contributions (and this is not theoretical, I've seen this in the wild).
(However, that's quite orthogonal to being a dick about making the source code that you must share available)
The originally envisioned distribution method, in fact, was "Send FSF a blank 9-track tape and they'll fill it and mail it back". Nor, obviously, does anything prevent someone who downloads this from Drive from mirroring it on GitHub or wherever.
This is arguably bad stewardship of a historically open source project. It's certainly not a license violation.
We said artificially delaying it for weeks or more is how they're violating it, not using Google Drive. It's also not provided in the preferred form for modification. It isn't in the form expected by the build system and causes it to not function in the way it did for their own builds.
Well to be fair you didn't say that in the linked tweet. Delay is bad too, but again the original spirit of the license was absolutely OK with the idea of weeks-long delays. Our modern world of instant github access has spoiled us, but courts are not going to see damages from just "we had to ship a month late" unless you can show something more specific.
It's required for providing timely security patches. It's not hard to show why that's important and Google engineers have written plenty about it for us to cite. We've managed to avoid delays for security patches so far through reverse engineering and other workarounds. Having to invest time from developers paid to work on GrapheneOS in reverse engineering compiled GPLv2 code shows Google is violating the GPLv2.
Delay is frustrating, but not a violation. Just because you reverse engineered GPLv2 code doesn’t make that in violation of GPLv2, many people have done that for reasons that rhyme with yours over the decades. GPLv2 also doesn’t have any clause about security patches or if you have a requirement for a certain timeline. GPLv2 isn’t a “you agree to not be an asshole” license.
(1) Niemand ist verpflichtet, deutsche Euro-Gedenkmünzen im Betrag von mehr als 200 Euro bei einer einzelnen Zahlung anzunehmen. Erfolgt eine einzelne Zahlung sowohl in Euro-Münzen als auch in deutschen Euro-Gedenkmünzen, ist niemand verpflichtet, mehr als 50 Münzen anzunehmen; dies gilt auch dann, wenn der Gesamtbetrag 200 Euro unterschreitet."
(1) No one is obliged to accept German commemorative euro coins totalling more than 200 euros in a single payment. If a single payment is made using both euro coins and German commemorative euro coins, no one is obliged to accept more than 50 coins; this also applies if the total amount is less than 200 euros."
To clarify given the subject at hand: German courts are 100% not going to find a Google Drive link to be disallowed by the GPLv2. That's literally about physical coins.
> This is explicitly about commemorative coins, not regular ones.
The formulation is not so easy to read (very common for German laws), but it includes also the regulations for normal coins:
"Erfolgt eine einzelne Zahlung sowohl in Euro-Münzen als auch in deutschen Euro-Gedenkmünzen, ist niemand verpflichtet, mehr als 50 Münzen anzunehmen; dies gilt auch dann, wenn der Gesamtbetrag 200 Euro unterschreitet."
"If a single payment is made using both euro coins and German commemorative euro coins, no one is obliged to accept more than 50 coins; this also applies if the total amount is less than 200 euros."
So, there exist two cases in which the vendor is not obliged to take more than 50 coins:
- The payment consists of both normal Euro coins and German commemorative euro coins
- The payment is less than EUR 200.
--
Independently, there does exist another source of law by which the vendor is not obliged to take more than 50 coins: Artikel 11 der EG-Verordnung Nr. 974/98 des Rates über die Einführung des Euro, EU-Amtsblatt L139 vom 11. Mai 1998:
"As from 1 January 2002, the participating Member States
shall issue coins denominated in euro or in cent and
complying with the denominations and technical specifications which the Council may lay down in accordance
with the second sentence of Article 105a(2) of the Treaty.
Without prejudice to Article 15, these coins shall be the
only coins which have the status of legal tender in all
these Member States. Except for the issuing authority and
for those persons specifically designated by the national
legislation of the issuing Member State, no party shall be
obliged to accept more than 50 coins in any single
payment."
Relevant part of this article:
"Except for the issuing authority and
for those persons specifically designated by the national
legislation of the issuing Member State, no party shall be
obliged to accept more than 50 coins in any single
payment."
It's not even that. Downstream projects host their own mirrors already, this is an annoying hoop to jump through for the maintainers (basically suck down a bunch of tarballs for every release, analogous to grabbing stuff from FTP sites back in the day), but not exactly a terrible hardship compared to the really very significant work of maintaining a large project.
There might be some merit to a claim that Google Drive isn't a medium customarily used for software distribution these days, but, yeah, it's definitely not paying a thousand+ cent bill in pennies, and I'm skeptical that it's a violation of the letter of the GPL.
> There might be some merit to a claim that Google Drive isn't a medium customarily used for software distribution these days
I suppose forcing a means to share the source code could have been too restrictive, but the GPL only speaks about the shape of the source code itself (it should be "the preferred form of the work for making modifications to it"), not how it is shared, so indeed, not a violation of the letter of the GPL I think.
It's like what we had in France and the Hadopi, which requested ISPs to share the IP addresses of people torrenting a defined set of files. One of them sent them printed on paper... (But the malicious compliance was cool in this case).
> I suppose forcing a means to share the source code could have been too restrictive, but the GPL only speaks about the shape of the source code itself (it should be "the preferred form of the work for making modifications to it"), not how it is shared...
With the greatest of respect, you've forgotten what the licenses say.
GPLv2: [0]
3. You may copy and distribute the Program (or a work based on it, under Section 2) in object code or executable form under the terms of Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable source code, which must be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three years, to give any third party, for a charge no more than your cost of physically performing source distribution, a complete machine-readable copy of the corresponding source code, to be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or,
...
GPLv3: [1]
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms of sections 4 and 5, provided that you also convey the machine-readable Corresponding Source under the terms of this License, in one of these ways:
a) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by the Corresponding Source fixed on a durable physical medium customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by a written offer [to convey the source code upon request]... on a durable physical medium customarily used for software interchange, for a price no more than your reasonable cost of physically performing this conveying of source, or (2) access to copy the Corresponding Source from a network server at no charge.
...
d) Convey the object code by offering access from a designated place (gratis or for a charge), and offer equivalent access to the Corresponding Source in the same way through the same place at no further charge. ...
e) Convey the object code using peer-to-peer transmission, provided you inform other peers where the object code and Corresponding Source of the work are being offered to the general public at no charge under subsection 6d.
This unambiguously speaks about the form in which the source code is shared. If the licenses didn't specify this, folks would be compliant with the letter of the license by shipping you a printout of the source code and everything you need to build it and charging you for both the labor to generate that enormous, heavy-ass printout and shipping and handling to get it to you. [2]
[2] To downvoters: Don't forget that OCR was decent even back in the 1990s... certainly good enough for a good-quality printout in a fixed-width font to be -strictly speaking- machine-readable, and it has only gotten better as time has wobbled on. If you don't believe my account of the history, go look up how Zimmerman exported copies of PGP back when it was considered an export-controlled munition.
People are getting way too bent out of shape over that "medium customarily used for software interchange" bit. It doesn't mean github. It doesn't mean "the medium I use most commonly".
Basically, if you think courts are going to be OK with interpreting "download from this FTP site" as acceptable but "download the same tarball from Drive" as unacceptable, you're fooling yourself.
Drive is fine, given the spirit of the license. It's merely inconvenient.
I hope you're not including me in "people". Remember that I said:
There might be some merit to a claim that Google Drive isn't a medium customarily used for software distribution these days, but ... I'm skeptical that it's a violation of the letter of the GPL.
I was quoting the text of the GPL to point out to jraph that it absolutely does restrict how source code is distributed to ensure that licensees are obligated to distribute in a format that's actually useful to the typical recipient, rather than permitting a licensee to ship a couple-hundred pounds of printouts and still be in compliance with the license.
Indeed, you are right, my phrasing "but the GPL only speaks about the shape of the source code itself" is somewhat wrong or at least incomplete. I should have been more careful. It does force some stuff about how to convey the corresponding source; and it seems the GPLv3 tries to close some loopholes or address some situations more explicitly. You cited the parts of the GPLv2 and GPLv3 I should have.
I stand by the position that all this doesn't seem very restrictive though. I don't think the GPL could have been without a risk of making some legitimate cases litigious or something.
> I stand by the position that all this doesn't seem very restrictive though.
Is your position that it's less restrictive than it needs to be?
If that's not your position, then I'm not at all sure why you're bringing this up. If that is your position, then I disagree with you. The entire point of the GPL is to require distributors to "share and share alike". It's not a "sue everyone into oblivion" license, it's a "don't be a fuckin asshole with this gift I gave you to use, inspect, and modify however you wish... pass it along to others under the same terms" license.
I think the GPL doesn't impose much on how one should be redistributing the source code.
I'm not sure I would like it to me more restrictive, and I completely agree with your reading (starting from "The entire point of the GPL...").
> If that's not your position, then I'm not at all sure why you're bringing this up.
My initial reply to you was me mostly agreeing with you: distributing via Google Drive is probably not a violation of the letter of the GPL. Making it a pain to get the source code is an obvious violation of its spirit though (your "don't be a fuckin asshole" point).
>> my phrasing ... is somewhat wrong
> It's completely wrong.
Well, what concrete restriction you see in the GPL about how to redistribute the source code, apart from "you must make it available in a reasonable way (and tell people they can get it, the GPLv3 is more explicit about this but Android doesn't have GPLv3 code AFAIK)?"
While I don't find any requirements on how timely the source distribution must be upon request, one can reasonably say that there must be a line between 1 nanosecond and 1 century.
Google can trivially provide it nearly instantly with no hardship. In fact, it's much harder for them to implement manual handling than automation. They have no justification for it beyond deliberately making it harder. It does have to be provided in a reasonable time or the license wouldn't work. What amount of time is reasonable is up to a court.
Google being incapable of timely handling of these requests is not believable. They're one of the largest tech companies in the world. They deliberately moved from a system without any need for manual handling of requests to requiring it with the clear goal of creating a hassle. By failing to provide it for long enough periods of time to cause tangible harm to people relying on it, they're failing to comply with the license.
Courts would most definitely make a distinction here. For instance, one century would mean "refusing to release the source code".
We should test how long it takes Google to release source code upon request. And whether it is 100%. I think we should test whether Google fulfils the GPL here. That's now a challenge.
We said artificially delaying it for weeks or more is how they're violating it, not using Google Drive. It's also not provided in the preferred form for modification. It isn't in the form expected by the build system and causes it to not function in the way it did for their own builds.
The basic requirement is whether the source code is available - and made available. Are you certain that Google's solution here is ensuring that the source code is easily made available? So many other projects just provide a wget-able link. Why does Google want to make it harder to obtain the source code than those other projects?
> Android has always been more source-open than “open source”.
And what exactly does that mean? I don't know what your words mean here. More source open than open source? Is that a tautology?
> Everything else is essentially all the work of Google and (to some extent) Samsung.
Is it GPLv2? If so then I fail to see why anyone should get higher rights. Everyone gets the same for GPLv2. That's the whole point. I don't understand your statements here.
> The basic requirement is whether the source code is available - and made available. Are you certain that Google's solution here is ensuring that the source code is easily made available?
The word “easily” sure did sneak into this sentence
I see your point, However the majority of Android devices have lots of closed source firmware/drivers, a large part of the Android OS doesn't run without them.
I'm refering to Bootloaders, TrustZone OS, Trusted Applications, then firmware for Bluetooth, Wifi, GPU, Sensors and power management. All of those are always closed source binary blobs running in the background.
> Android has always been more source-open than “open source”.
I hate Google as much as the next person, and the only way with those companies would be to fine them heavily, quickly and systematically.
But it is important that we are clear about what Google does bad and what it does well. And I believe that understanding what "open source" means is important when one wants to talk about a violation of an open source licence.
Open source does not mean that they take contributions. Many firmwares used by Android are not open source, but AOSP is open source. It is licenced under a permissive licence (Apache 2, I think for everything), which makes it open source, period.
I think if you watch video of a massive meatpacking plant, you'll pretty quickly realize how hard it would be to track. At the commodity scale, basically no one in the supply chain other than the rancher seems to care where the meat comes from. And the brands (Tyson, Smithfield, etc.) only want consumers to care about the logo on the box.
Literally the canonical example of capitalism-as-meat-grinder.
Thanks for the correction. I think the issue here is that none of this information is public (like I can’t call the company and ask where lot 11956 came from), so it’s impossible to know as a consumer.
Tyson in particular. Child labor? Had issues with that. Institutionally approved use of undocumented workers? Check.
And a willingness to throw both under the bus when OSHA comes knocking because they also have a large number of workplace safety incidents. Or when the workers complain about such conditions, a little call to DHS and all of a sudden ICE shows up for a raid to pick up 900 undocumented workers, but when a reporter asks officials about the fact that Tyson was giving them written instructions on how to stay under the radar for banking, tax, immigration purposes? "That's not something that's in the scope of our investigations here."
So you're saying that this type of regulation would force the food industry to start caring about whether their food comes from? That sounds like an argument in favor of it, not against it
trying to convert a massive meatpacking plan isn't step 1. it's too big a problem. that's top down. probably easier bottom up -- have a processing plant that only sources from ranchers that grass feed/finish, pasture raise, and slowly subsidize other ranchers to stop sending their product to massive meatpacking plants.
Seriously, what is the deal with Pip! I was experimenting with childrens' fiction more than a year ago and 9 times out of 10 you'd get a character named Pip unless you were specific in the prompt not to do so.
Well, they could have used a less generic and misleading name (it is not very open, as noted in the article). OpenAI only really have themselves to blame here.
Plenty of companies use generic words for their name, and they still get trademarks.
American Airlines for example is indeed just an American airline. The Container Store, Vision Center, General Motors, International business machines (IBM), the list goes on.
Even Microsoft is just a contraction of their original product, microcomputer software.
I understood it more in the line of preventing a company from naming itself "Low sugar" and then blocking other companies from adding the words "Low sugar" to their packaging. Same thing with OpenAI, another company should be free to create an AI that's fully open and tag it as "Open AI" without fearing legal problems with OpenAI.
I would say there is nothing descriptive about calling something American. America doesn't even refer to a single country, does it operate out of Brazil? Would that even be relevant information to anyone using an Airline?
Exactly. Apple can register Apple because they don't sell apples so it's not misleading. OpenAI can't register OpenAI because they make ai but it's not open. They could call themselves Peaches, OpenWombat or ClosedAI and there wouldn't be any issues because those wouldn't be misleading.
> OpenAI can't register OpenAI because they make ai but it's not open
That's not the reason they can't. They can't register the trademark because it's a descriptive one.
If I try to trademark "hacker forum", an EU trademark officer will reject it not because my website doesn't have hackers on it, but because it's descriptive and prevents others from starting hacker forums.
So
> They could call themselves... ClosedAI
is also incorrect, because it's descriptive as well.
You know, if it was someone offering a truly open (weights + training data) and available model running on consumer hardware in a privacy sandbox, I would welcome that "harm".
Ah yes, chosing a name that transports openness and transparency when the opposite is the case, and complaining about not being able to register that name as a trademark, which will cause financial harm the said company -- but somehow there's still people to spin it the other way around so it harms consumers now, therefore it was a bad decision.
You just don't like OpenAI and are for anything that hurts them, without thinking through the consequences.
What will harm consumers is the scammy "OpenAI" chat app that I can now legally upload to app stores in the EU, in hopes of tricking people into thinking it's a genuine app.
It seems pretty dubious that there are a lot of people who know the name of the company behind ChatGPT but not able to recognize that the name of their chat isn't "OpenAI chat"
Seriously though if you are letting agents do whatever they want without a PR process that requires hardware authentication or proof of presence, you are putting your code and your org at high risk.
> want without a PR process that requires hardware authentication or proof of presence
Just curious, what do you use for this?
I built OTP Guard [1] a few years ago for exactly this problem, although I haven't seen any alternatives in the space. Does GitHub have something built-in now?
The original framing was more "local malware compromising your GitHub account" ... it never occurred to me that the malware could be a LLM. I really should update the page.
0) agent gets its own separate git user and ssh key, separate from mine
1) branch protection rules on main, only I can approve merges into main
2) any other ssh key uses (interactive login, direct git access, etc.) are ed25519-sk keys and require a touch on yubikey.
TBH, the biggest hole is that it can be unclear exactly what process is requesting a touch on the yubikey. Apple has a head start here because they can lock down the TouchID UX relatively well, but unfortunately they don’t seem to care about building a polished developer experience for 2FA on sensitive tasks.
They are probably waiting for someone else to build the right solution and then copy/steal it.
