HN user

aftbit

6,381 karma

https://alloscomp.com/ https://github.com/lachesis

Software developer in Ohio

I've worked at: 2009-12 L3 CE, 2012-14 Etegent, 2014-15 Dropbox, 2016-24 Zinc, 2025 Atomic Canyon

Email: ems@(my-username).com

[ my public key: https://keybase.io/aftbit; my proof: https://keybase.io/aftbit/sigs/-UWz7QJtEJD409eCqmyDiAP03IVBDKY-mKh5akhqi60 ]

Posts4
Comments1,882
View on HN

Yeah... that's been my past experience with USB docks, at least any with more than one slot. I've never had great luck with them. "Can recover from power outage without being touched" is a key requirement for my NAS so I'll give that one a pass and stick with my "SATA drives directly attached to a SAS controller" strategy for now. Thanks for the reply.

As for the 24/7 use, yeah so be it. The I in RAID stands for Inexpensive. If they fail after 10 years at 24/7, so be it. I have drive level redundancy and frequent offsite backups of anything critical.

Neat blog! I was intrigued by this bullet point mentioned in passing:

my four hard drive USB 3.2 ZFS raidz2 array with four 24 TB drives

Can you speak more about this? Which USB array did you choose? How well does it work? I've been slowly planning a transition away from my power-hungry surplus enterprise gear in the 19" rack towards a smaller, quieter, lower power setup ... but storage is the real kicker right now. I have a 12x18TB array in raidz2 built into a 1U NAS case, and I just can't quite figure out a better way to package something like that. I would need three USB arrays if I want to reuse the existing drives, which I think I do given how expensive storage is today.

SpaceX-the-space-launch-business is doing amazing things, which is part of what makes it so sad that the financials and buzz is all for SpaceX-the-enterprise-AI-business.

Their S-1 projects $28 trillion TAM, with 93% of that attributed to AI, ~6% to connectivity via Starlink, and only ~1% for space launch capability.

Of their $20 billion CapEx spent in 2025, 60% was spent towards AI data-centers, 20% was spent towards Starlink satellites, and ~20% was spent towards actually building Starship and Falcon 9. That's gone up to nearly 80% towards AI in Q1 2026.

Of course, "AI" makes up only ~20% of their current revenues.

So... this company is priced as an AI startup but it's actually a satellite internet and space launch business.... or is it? Shame to see the coolest thing that Elon did be eaten by the AI monster.

https://www.sec.gov/Archives/edgar/data/1181412/000162828026...

Doesn't matter if the business fundamentals are sound.

The business fundamentals are rarely sound for modern IPOs, especially anything Elon adjacent. His companies are just as bad as crypto token sales in terms of their hype. Heck, some of the stock price appreciation of Tesla _was_ driven by their ownership of crypto for a year or two.

And yet, because these experimental owner-built aircraft are free from many of the burdensome certification requirements, they're also free to use newer and better tech at a lower cost.

Who is safer? The person who "self-built" (but actually allowed the trained factory folks to build) a Kitfox with a full glass cockpit, modern engine, and dual FADEC, or the person who buys a 1972 Cessna Skyhawk with steam gauges?

As I understand it, a big part of the problem is that the general aviation market is too small to justify the expense of getting new designs certified. Also, it seems baked into the regulations that the old design was "good enough", so it still is today, even though technology has moved on.

When people are building experimental class piston airplanes with better features, reliability, and performance than the certified world can muster at half the price, you know something has gone wrong.

We're working on it, slowly but surely. This has been a known problem for a long time, at least since 2009 if not earlier. GAMI's G100UL was approved as a 100LL replacement in 2022, but there is still some controversy.

Avweb has done a good series of videos on this. There were some real engineering concerns and some typical aviation conservative decision making. But really, it's a tragedy of the FAA fumbling the ball for decades at this point.

https://www.youtube.com/watch?v=9F-WngVMJBQ

https://www.youtube.com/watch?v=Mvse4Xhzwuk

https://avweb.com/features/unleaded-avgas-airport-playbook-l...

https://avweb.com/aviation-news/judge-denies-g100ul-motion/

https://g100ul.com/

I mostly agree in my 2024 Ioniq 5, but not in my 2019 Subaru Outback. You can definitely override the lane keep if you have a firm grip and are ready for it, but it tries to throw me off the road often enough that I don't use it anymore.

The scariest was when I had to swerve into another lane to avoid some trash that was sticking into the road from the highway. It tried to force me back into it twice! Luckily I was ready but it gave me a fright for sure.

Not sure about the systems on cars in the EU, but I got a loaner 2025 Hyundai Tuscon when my EV was in the shop. It had some driver attention monitoring feature with a camera above the steering wheel staring me in the eyes. I covered it with a piece of black electrical tape. It popped a little warning on the main display (IIRC, a crossed out eye, but maybe I'm confusing with Subaru Eyesight) when the car first started up, showing that the camera wasn't working, then proceeded to be silent for the rest of the drive.

I dunno if that'll fly going forward. I know I'll test it in every new car with this feature that I test drive though!

Computers are pretty fast these days. It depends on exactly what kind of general purpose use you're talking about. Trying to push 10 Gbps through a slow old machine? You're doomed. Trying to run SSH, browse the web, or even watch Netflix over a tunnel? Probably fine to send the packet up and down the stack an extra time or two.

They're getting better though. The first gen touchscreens were tiny and unreliable. The one in my 2024 Ioniq 5 is pretty decent. I am really glad I still have physical AC controls though, even if they're capacitive.

Touchscreens are modal. If I want to control the climate, I first have to press the Climate capacitive button or scroll through the screen to find climate. That takes my hands off the wheel and my eyes off the road for longer than just tapping the fan-up button.

As for the cost, I will _happily_ pay the $100 more to have a more premium and tuned interior. Heck, I chose to step up an entire trim model to the top of the line trim just for the fancy LCD screen mirror. I'd happily pay extra for better buttons.

IMO touch screens are great for rarely used features, but anything that gets clicked on most drives should be a dedicated touch point (capacitive button, physical button, steering wheel control, whatever).

Give me multifunction displays from aviation. Touch screen in the middle, rows of modal buttons along the bottom and left side. You can use muscle memory to find the button.

But ... Rayfish creates a layer 3 network too. It assigns IP addresses to each node. I don't really see why it matters from this point of view whether it uses Iroh or something else to provide its backhaul. That's like saying Wireguard works at layer 4 because it uses UDP. The whole point of VPNs is to offer some kind of tunneling over higher level protocols.