HN Simulatornew | past | comments | lists | submit | als0's commentslogin

How to say that you worked at Microsoft without saying it

Ahahahaha, nope.

For the 1990s I was wildly anti-microsoft person, we were an OS/2 home for big chunk of it, WordPerfect over Word, etc.

Over time I learnt just what kind of advantages beyond the popularly villified ones Microsoft ended up having in sales, and Windows 2000 was really a huge blow to competition in many ways, because it was first product that integrated it to this level.


Also related is Propolis from Oxide who worked on live migration with Bhyve https://github.com/oxidecomputer/propolis

Hi, Original propolis architect here, and primary engineer for much of the effort porting bhyve to illumos.

Quite a bit was changed between "illumos bhyve" and FreeBSD upstream. We in illumos decided to take a slightly different approach in how we approached the machinery for saving/restoring the state of devices emulated in-kernel, which is a necessary part of doing such a live migration. FreeBSD was first to add that functionality, but we (I) felt it exposed too much in the way of implementation details from that device emulation, rather than a more abstract model of the emulated hardware. It was at that point that illumos "soft-forked" the kernel VMM component so we could take a different approach. While the two kernel VMM components are very similar, and much of the bhyve userspace emulation (a counterpart to propolis) will run on either OS, there are some key differences around instruction emulation and those kernel-emulated devices which prevent them from being drop-in replacements for each other.

All that said, I'm happy to see FreeBSD making progress on this front. They've been a friendly peer in this space, and both projects benefit from efforts in this area.


Not really familiar with bhyve or even illumos in the slightest. But reading this comment makes me a bit sad in a way. These types of conversations and comments used to be common where I worked just few years ago. Now no one even cares how something is implemented let alone a "I felt it exposed too much in the way of implementation details from that device emulation, rather than a more abstract model of the emulated hardware". The "good" ones today are looking exclusively at external properties rather than internal ones. Thank you for the comment.

> Now no one even cares how something is implemented ...

> The "good" ones today are looking exclusively at external properties rather than internal ones.

To provide a little additional context here: This effectively was an external property of the system (just to the userspace VMM). One of the significant challenges for those implementing a feature live migration is how you'll deal with multiple version of the software (older and newer) interacting in the future. A primary use case for live migration is moving workloads around to permit software updates, making that multiple version issue even "worse".

Abstracting the implementation details of those in-kernel emulated devices was helpful in two ways. Firstly, it reduced (hopefully) the chances that a later change to the internal implementation would be visible in the exported/imported state. Secondly, that the act itself of doing that state abstraction required a much more intentional and thorough examination of the emulation logic itself, resulting in a cleaner and simpler artifact.

So while it may seem like it was driven by a desire to take pride in the craft (undeniably _a_ factor), the main motivation was avoiding a versioning mess in the future.

> Thank you for the comment.

My pleasure!


By your storing of a more abstract representation of the emulated device state, could you potentially live migrate to tape or object storage?

I feel you. Good programmers still care about those things, but it turns out there are a whole lot of people who never cared about the craft and were just in the profession because it was lucrative. It's a shame.

Where can I buy one?

Drop me an email (see bio)

Soon you’ll “Tell claude” to write your posts


I was waiting for that punchline.


You wanted to make a humoristic slippery slope effect, but I'm afraid they pulled the rug out from under you and actually suggested just that.


Who is cappelikan? Why should I trust them?


That "mainline" is only installer tool:

https://github.com/cappelikan/mainline

You don't need to use that mainline installer.

You can download mainline kernel .deb files directly from Ubuntu:

https://kernel.ubuntu.com/mainline/

https://kernel.ubuntu.com/mainline/v7.3.1/


I found here Canonical commenting about status of new kernels:

https://news.ycombinator.com/item?id=50348688


Why? Do you think frontier AI will never generate enough profit to justify its costs? The market seems to move on vibes rather than logic. I can imagine one of those company failing and the other seizing the market.


while what you say is true, trillions of dollars and thousands of people invested in something that ultimately leads to a few companies from china doing the same 10x cheaper is a vibe that can go so far


I know that there are a few regulated industries that will go out of their way to not have any prod data ever exposed outside of the United States.

"China can do it cheaper" applies well to your personal projects. It doesn't apply well when you handle things like trade secrets, healthcare, or financial data.


> It doesn't apply well when you handle things like trade secrets, healthcare, or financial data.

OpneAI stole valuable secrets from their customers Buckmaster and Alpöge, why should a company expect them not to do the same with their trade secrets?


I'm looking at the wiki here [0]

What was the nature of Buckmaster & Alpöge's business relationship with OpenAI?

When we started using Claude at my work, it wasn't just with Anthropic-hosted instances of the service. As I understand it, we use a copy hosted on a GCP Vertex AI instance. It'd be weird if Anthropic ended up with our data.

If Buckmaster & Alpöge had some sort of agreement with OpenAI that their inputs would not be used to train OpenAI's models or otherwise generate revenue for the company, yeah, that's theft.

If Buckmaster & Alpöge were using the same OpenAI-hosted services as everyone else with a standard EULA that effectively says "we get to look at your data and do what we want with it", well, then I don't know what they expected.

[0] https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_priority...


I promise you China can do it cheaper is good enough for many corporates that have neutral views at best on the labs

These are open models, go speak to corporates see what they’re up to

Also the healthcare bit is funny. Go ask the healthcare bigs and their intermediates what they’re doing today, hint it’s India

American corporates have zero loyalty to America or Americans, and they love bypassing laws, attempting regulatory capture on American corps is braindead and shows a weak hand, as they speak all this bluster about 2-3T valuations for an easily reproducible technology


> Also the healthcare bit is funny. Go ask the healthcare bigs and their intermediates what they’re doing today, hint it’s India

Funny you should mention that.

I am a team lead in a healthcare IT company. On my team are two people in India and one person (besides me) in the US.

These aren't contractors. They're full-time employees. Vetted, interviewed, subject to corporate policies like everyone here in the US, given a VPN to connect to things with.

Wanna guess what they are absolutely forbidden from doing by the contract that our clients signed, and by extension, what the people located in the US spend a lot of time doing that they otherwise wouldn't have to?

> American corporates have zero loyalty to America or Americans, and they love bypassing laws, attempting regulatory capture on American corps is braindead and shows a weak hand, as they speak all this bluster about 2-3T valuations for an easily reproducible technology

No arguments there, they should be taught the meaning of loyalty.


huh? assuming you're in america right now, the device you're using to publish that comment has been built in china or had most of it's components anyways, which are from some very trade secret-sensitive companies


Tangentially: The costs could be justified even if it doesn't turn into profit, no? Because if the price of good, robust software plummets, then that itself would be a a solid return on investment for the world.


> Do you think frontier AI will never generate enough profit to justify its costs?

I don't know, but I'm not sure it matters. Anthropic and OpenAI still have such large market shares only because their services are heavily subsidized. This can't go on forever.

At some point, they will have to stop burning investor money and start hiking prices. At that point, all of their competitors will look even better than they do now. So I don't see how these two companies in particular can generate enough profit to justify their costs.


Correct and before some idiot comes in and says ‘high gross profit margin’

Yeah shut up idiot - they need to raise prices to cover operating expenditures and reinvestment that stays inline with the valuation. If it implodes they’ll have trouble raising money later and they’ll need to lay off people - aka on bad terms.


They and their partners can’t pay off the debt to buy top-of-the-line inference hardware before it becomes obsolete and requires more debt to buy the next generation of top-of-the-line inference hardware.

This cycle will continue every 3-5 years. Given the supply for this hardware can’t keep up with the demand, they’ll be expected to pay more with each cycle.


Everything aside from apps with SafetyNet dependencies. Banking apps come to mind.


Also, a lot of games. This is actually why I stopped playing Pokemon Go, and never picked up a worse gacha game habit haha


Pokémon Go had a good reason - it turned out to be foreign intelligence gathering operation and couldn't risk being corrupted with false data.


What? Niantic (Google) admitted they were training spatial AI models. I have not seen evidence that the data was a "foreign intelligence-gathering operation" and I would welcome a source.


Which models for who?


Niantic plans a “Large Geospatial Model” trained on Pokémon Go player data | https://nianticlabs.com/news/largegeospatialmodel | https://news.ycombinator.com/item?id=42187494

430 points | 2 years ago | 338 comments


Soo many more things than banking apps at this point. Things as normal as looking at train schedules, grocery store catalogues, ordering a taxi... sometimes have SafetyNet DRM.

When you publish apps on Google Play, the Play Developer Console even nags you to use harder obfuscation and implement Google's DRM to, allegedly, protect yourself and your users.

Can the EU finally make this shit illegal? Oh wait, they're pusing for mandatory DRM too... Fuck.


Banks want your money so try it anyway. Not all banks try to drive away their customers by preventing them from accessing their accounts.


Sure, but who would use a banking app when the web site is right there and has more features and generally works better?


These days many banks only allows app for "enhanced safety"


Fortunately, there are (at least for now) many, many banks to choose from, and you can tell the app-only ones to suck it. There's also almost no switching cost. The last time I changed banks, it was mostly done in an afternoon.


I've never heard of such a thing but that's definitely not a company I would do business with. If my current bank tried that shit I'd be gone.


FWIW you can use also convert ReStructuredText (RST) to LaTeX. RST is very similar to Markdown, and it provides useful features that are missing. It also renders consistently across different viewers (presumably due to a tight specification). If you're working a lot with tables then you'll really enjoy the ease of the "list table" syntax.

I'd say RST is suitable for many types of documentation but I'm not convinced that it's suitable for conference/workshop submissions.


I've always thought CSP as a robust design pattern where you have no shared state between components and they must communicate with each other using message passing. It also requires synchronous communication (rendezvous-style). If you follow those rules you can have a pretty robust system. Aside from these abstract rules, CSP has more formal research (algebra) but I'm not sure if there are any decent tools available.

TLA gives you a full toolbox and in theory can model whatever you can express. That's very different from a design pattern.


Why decades?


Need risk aversion to reset after risk was raised temporarily. Humans are cautious creatures


Less risk aversion and more about momentum. You remove billions in funding, people leave the industry that took decades to obtain the education to be functional in, and at the same time now there are less people to train those who will replace them, and you still lack the funding.

If tomorrow you just inject 50% more funding, it doesn't mean 50% more science gets done tomorrow.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: