retrovm

You know I think that cloudflare blog post is a great example of the engineering approach behind boringssl. It's optimized for actual workloads, where you decrypt or authenticate a short message and then move on to different activities, and engaging AVX512 doesn't actually pay off in reality. OpenSSL is optimized to produce the biggest number from `openssl speed` so of course in that light it makes perfect sense to enable AVX512. But if you're trying to use these libraries in realistic workloads you will begin to appreciate the boringssl approach.

No, it's just a lot smaller. A 12-foot-wide tunnel is, shockingly, much cheaper than a 21-foot-wide tunnel. It also has essentially zero capacity, much like a hyperloop. If you want to punch useless holes in the earth, Elon Musk is your man without a doubt.

San Diego actually passed an ordinance requiring scooters to geofence speed limits of 8MPH in many areas. See https://www.kpbs.org/news/2019/jul/01/dockless-scooter-regul...

It's a great idea and they should absolutely expand this regulation to cars.

E-scooters instantly dominated the market share of shared micromobility when they debuted in the USA in 2018. https://usa.streetsblog.org/wp-content/uploads/sites/5/2019/...

The evidence in favor of wearing a helmet while in a car is just as strong, but nobody is up here advocating for that.

By the way, helmets on bicyclists tend to make car drivers behave like even bigger jackholes than usual. https://psyarxiv.com/nxw2k "... public health research might be best served by shifting focus to risk elimination rather than harm mitigation."

I hate that HN is a place where the ignorant simply flaunt their ignorance instead of trying to inform themselves.

"According to the Velofollies survey as well as other sources last year’s [2018] e-bike sales stood at slightly over a quarter million units; some 251,500 to be more precise. This accounts for a big 14.3 percent growth compared to the total of the year before. It makes electric bicycles by far the biggest category of the Belgian market with a market share that stands at close to 50 percent."

https://www.bike-eu.com/sales-trends/nieuws/2019/07/market-r...

I have a suspicion that our industry harbors an above-average number of bicyclists.

You're welcome to wear whatever you want while riding your illegal whatever-you-call-it. Don't scold the rest of us.

You need a special license for mopeds in many jurisdictions, they weigh 100+ pounds, double what an e-bike weighs, and because of the fuel system you can't really just carry it up the stairs and leave it in your kitchen at night.

Helmets are for sports. Ebikes have a ~50% market share in Belgium where approximately nobody wears a helmet ever. It's really not an issue.

I'm not sure what the point is, actually. The variety of things that can go wrong with them is astonishing. For one thing, among many others, BMC cards will power-cap the max clock speed of CPUs when a machine is running on only one PSU, which can cause a degradation that's worse than if the machine had just halted. There are a zillion other edge cases like that.

It's adorable that you think batteries and dual PSUs are something that makes a node more reliable, rather than less reliable.

"no real need to make your implementation more complicated than a single array"

That's our industry in a nutshell. Our computers, instead of becoming more capable over time, can barely keep pace with the increasing naivety of our programmers.

POSIX threads can be individually named. They inherit the name of their creator if you don't set one.

That presupposes that you can start with a big pile of bad APIs and somehow incrementally approach something worth having. That's not consistent with my experience. In my experience the innermost APIs, the ones that get written down first, are effectively cast in stone and dictate the quality of the whole product, forever. The first draft of a system is the one that should have all its APIs worked out with a pencil before any code is written. You get these accidental complexity explosions precisely because someone thought it was easy and obvious to write a function returning std::vector<string> (or whatever) and didn't consider from the caller's point of view that making the function take an output iterator might be less complex.

Well, estimates I can google up at the moment put the Pb-acid worldwide market at ~$42bn/year and the Li-ion market at ~$40bn/year. The latter is growing much faster but the former is also growing.

I think this is an important point. The lithium-ion battery market is _still_ smaller than the lead-acid battery market. This business is still ramping up.

Being a good-enough CPU for terrible software is a bit of a tragic conclusion, in my view.

Beats me but I happen to have a fleet of HP SATA (not SAS) drives and they just crossed this boundary, 32813 power on hours typical. I guess if their SATA firmware had this bug I'd be having a bad week.

If you actually read the article most of the statements of fact are about the omnibox autocomplete system, and then they use innuendo to imply some things about search engine ranking. But these are two completely separate systems, and it makes sense that a system that is literally telling you what to type is more sensitive than the search result ranking. It is not a flaw of Google that it won't suggest "is hillary clinton still controlled by the jews" when you type "is hillary clinton". If it was just a big trie of what everyone typed, it would be completely dominated by 4chan troll bots.

2018, FWIW.

I seriously doubt that illumos has the edac sorted out on newer parts. That's why i recommend running it in a VM.

I use it daily but I haven't "used it" in about 6 years because it just sits there in the corner serving without me having to touch it. When I needed docs I used to just refer to Oracle's.

It does not run on Linux, because parts of it are implemented in the opensolaris kernel. It is dramatically better than Samba.

Linux chipkill ECC support is a crapshoot. Intel redesigned it all for Skylake and later CPUs and Linux still hasn't caught up with the changes. Some of it mostly works on pre-Skylake, but often doesn't. I've seen Linux killing processes that have a bad physical page mapped, but the process just gets killed in a loop because the kernel is too stupid to actually take the page offline.

The hardware support is thin, so it makes sense to run it in a VM. I run mine inside VMWare ESXi, which has excellent hardware support. I give all the storage to the OpenSolaris VM, which in turn re-exports iSCSI volumes that are used by the other VMs. Solaris's iSCSI daemon is rock-solid, as is their CIFS server. It's miles ahead of the network storage daemons available for Linux. And of course it's all backed by ZFS.

Are you joking? The car is driving down the wrong side of the road.

Tremendous amounts of ciphertext?

That's a different problem though. The fact that Facebook employees access your private data is not because of insider risk being taken unseriously. It's because looking at and distributing your private data is a core function for that company. They don't consider it a flaw.

The centralization of infrastructure has really changed the insider risk profile at that company. A decade ago it was common for a service to own their hardware and to have broad authority over those machines but not so much these days. Nobody wants to own their hardware, everything is uniform and it's easier to spot deviants that way.

Of course many of the changes around that time were in response to the breach by the Chinese government, not only in response to embarrassing privacy incidents. Later improvements came about because of things Snowden published.