Zimmie

Dan Kaminsky once said I know how computers work.

Pronouns: he/him

Last Notes

2025-06-17 20:51:51 UTC

- reply

I’d be okay with cops disabling bodycams as long as anything which happens while the bodycam is off is legally the cop’s fault.
Somebody dies in custody while the cop’s bodycam is off? Cop is automatically guilty of murder 2. Somebody says money disappeared? Cop stole it.
Make them *want* to keep their bodycams on and in good repair.
Of course, getting there seems about as likely as getting rid of qualified immunity.

2025-06-16 18:20:43 UTC

- reply

There are tons from various brands (Dell, EMC, HPE, IBM, Netapp, Promise, Sun, Supermicro). A 2U enclosure fits 12 3.5” drives. 3U fits 15 vertical or 16 horizontal.
4U gets into Backblaze Storage Pod territory. Up to 60 drives per enclosure (sometimes more with weird layouts).

2025-06-04 19:45:22 UTC

- reply

As opposed to the cost of dealing with photovoltaic waste, which we offload onto developing nations. Watt for watt, they produce 2-3 orders of magnitude more waste in both mass and volume than nuclear power does. And the waste from photovoltaic panels doesn’t get less dangerous with time.
Or the cost of dealing with wind turbine waste, which is likely to never be recyclable (turbine blades are fiber-reinforced composites). Right now, the best disposal method is to grind the blades into chips and burn them for process heat. Less toxic than photovoltaic waste, and the risks to workers (like silicosis from grinding the blades into chips) are mostly manageable.
The only power generation methods which makes less dangerous waste than nuclear are geothermal or hydroelectric, both of which depend heavily on geography.

2025-05-12 18:43:18 UTC

- reply

Absolutely!
Note that you can also get used/refurbished devices to save some money, or even build a storage server for yourself to meet your exact requirements.
I’m also a big fan of “refurbished” server drives for primary storage. Modern high-capacity hard drives are filled with helium instead of air, and they are expected to last millions of hours.
There’s a technology called SMR which is used to increase capacity at a performance cost most find unacceptable. It’s mostly used on consumer drives rather than server drives, but you should still look for datasheets on drives and check for it. CMR or PMR are good.

2025-02-16 01:52:52 UTC

- reply

Please try to enjoy all casting equally.

2025-01-30 20:48:56 UTC

- reply

Apparently the civilian plane was banking and couldn’t see the military helicopter, and the helicopter thought they saw the civilian plane but they were looking at some other civilian plane.
Not sure if it’s the case here, but military flights routinely have ADS-B disabled. Something something national security something.
The collision was at just 91m altitude. TCAS generally mutes advisories below 300m, right? Otherwise, aircraft would get constant “Climb! Climb!” from reflections and ground clutter as they land.

2024-12-17 03:24:58 UTC

- reply

Yeah, I drive a *tiny* car, and I almost always let the cars to my sides go into the intersection well before I do.

2024-11-29 17:44:55 UTC

- reply

Excuse me, but I follow for corn content.
Post cob!

2024-11-27 19:03:15 UTC

- reply

Fortunately, SATAv3.3 (irony alert!) released in 2016-02 and officially declared that pin is now for power disable. There’s still a risk somebody followed the old spec, but it’s getting lower over time. Backplanes and power supplies built in the last few years generally don’t put 3.3v on the pin.
What really irritates me is they could just as easily have said pulling the pin low reboots the drive. Then drives would just work in whatever enclosure.

2024-11-27 18:43:55 UTC

- reply

Yep! Modern drive heads’ fly height is somewhere around 500 picometers. The van der Waals diameter of a helium atom is about 140 picometers, so the head basically rides on top of a bed of helium atoms which act like ball bearings.

2024-11-27 16:52:45 UTC

- reply

Hard drives still need *something* to keep the head from physically touching the platter. With hard vacuum, they would vacuum cement together in a fairly short time.

2024-11-27 14:54:45 UTC

- reply

draid is a really cool new-ish feature which is *great* for big deployments like that. It gives you the simplicity of a single vdev (you don’t have to care about the *wrong* three drives failing at once), the IOPS of multiple vdevs, but the headline feature is the distributed spares.
The problem with resilvering in general is that it’s limited to the performance of a single drive: the one you’re resilvering to. By distributing the spare capacity between all of the drives in the vdev, you can resilver *much* more quickly and regain your fault tolerance. You still need to replace the failed drive eventually, but when you do so, you’re already at your design fault tolerance, so it mostly doesn’t matter how long that takes.

2024-11-26 17:59:03 UTC

- reply

With 44 drives and that weird setup, I’d probably use draid2:8d:44c:4s. You lose four drives off the top as distributed spares, then the remaining 40 are split into groups of 8 data plus 2 fault tolerance for 4 groups x 8 drives per group x 12 TB per drive => 384 TB usable before compression.

2024-11-22 23:44:59 UTC

- reply

Yep, it’s excellent. Have you seen Constellation? Also extremely good.
Silo’s second season just started, and it’s just about perfect.

2024-11-18 17:15:19 UTC

- reply

This kind of shot is almost always done with a pinhole camera. You leave the camera in place for a year. Each day, you go out and uncover the pinhole at the same time for a few seconds.

2024-09-29 19:26:36 UTC

- reply

Agreed, adjusting the timesharing to support more clients per tower at the expense of latency and throughput would be a a decent step. Limit each terminal to maybe 2 megabits of data per second, and you have time slots to work with more terminals.
The downside is still that cell connectivity depends on centralized infrastructure. 24 hours of power is massive overkill in almost all situations, yet it’s also far too little for a lot of the natural disasters we’ve been seeing. It can take multiple weeks for power to be restored in some areas, but it’s impractical to require two weeks of backup power at every cell site, especially when 90% of cell sites are likely to go years without ever tapping the backup power. At least since cell towers are pretty tall they can use non-battery power storage (such as a motor connected to a very large concrete block).