scott00

I'll add that you should look for the PLUS skus, ie SFN-8522-PLUS. They come with all the fun stuff.

Might also want to look at Packet. They offer bare metal cloud servers by the hour, some of which have Mellanox NICs in them[0]. I'm less familiar with Mellanox than SolarFlare, but I think they have a similar feature set.

[0] https://www.packet.com/cloud/servers/c3-small/

This is likely not true. You can't just sign up for FINRA exams as an individual; an eligible entity (broker-dealer or investment advisor) has to sign you up. And even if you have passed the relevant exam recently enough, you're not considered licensed unless you're working for an eligible entity who has submitted the appropriate paperwork to claim you as a registered representative.

The easiest path this opens for someone not in the securities industry is to become a state registered investment advisor. In my home state of Illinois, that would require paying $400 a year to the state and $150 a year to FINRA, as well as subject you to a number of non-trivial regulatory requirements.

The NSA doesn't need to know a backdoor exists in order to worry about one. The ME is infrequently exercised code with a large attack surface and highly privileged access. If you are security conscious and don't need the functionality it's quite logical to want to turn it off, whether you're the NSA or anybody else.

I use jumpcloud (https://jumpcloud.com). Works well, free for small teams. Not open.

I was also annoyed by that. However he didn't include interest in his repayment numbers either, and the two errors roughly offset.

So then how do you manage the secret that authenticates an application's identity? And what good is the logging if after an application has the secret it can do whatever it wants with it?

> fsync() is required to ensure that the data is on disk, though I've seen systems lie about that too.

Virtually all SSDs lie about this to a certain extent. They'll indicate a write is complete when it's in their cache. The cache is RAM; as long as the power stays on, the write will appear to have succeeded to readers, but if the power goes goes off before the drive flushes the cache, the write will disappear. The exception is high end enterprise drives, which come with a capacitor that stores enough power to flush the cache to persistent storage in the event of power loss.

The goal is to prevent garbage collection from happening during fast path execution. So the program doesn't need to be literally allocation free, just free of heap allocations during the steady state of the program. So heap allocations during the initialization phase are okay, as are stack allocations at any time. And the no heap allocations during steady state isn't as bad as it sounds, as using an object pool that has preallocated objects in it is usually a reasonable replacement.

Is profiling actually a good answer to that question? Profiling is very useful for improving the performance of software under a particular workload, but unless you have very good non-anonymized telemetry anything other than trying to get more information about the customer's situation is going to be a wild goose chase. I would definitely start with talking to the customer to get more details, and ideally observe the customer using the app before I got anywhere near a profiler.

Some useful context: Zoom also runs 17 of their own data centers in addition to using AWS and Azure[0]. This deal is to run their "expansion" on OCI. Hard to say what that means exactly, but it doesn't sound to me like Zoom is making a big bet on OCI.

[0] https://www.cnbc.com/2020/03/18/zoom-cfo-explains-how-the-co...

The confidence intervals are wide enough it wouldn't be very useful. If you sample 1000 people and get 0 positives, the 95% confidence interval is 0% - 0.36%. If you're the governor of Illinois, with a population of 12.67 million people, that upper bound equates to 45,612 people: far, far too many to handle with contact tracing and isolation. So you need to look to other metrics to make the decision about whether to issue or revoke a stay at home order.

So what are they looking at? Basically, if they have any evidence of community transmission, they need to issue the order or keep it in place. That means that if there are too many symptomatic people to even attempt contact tracing, you're not even close. Once you're low enough you can do contact tracing, you wait until you have a handful at most of symptomatic people you can't trace back to a known case. Until then, all you can do is keep the stay at home order in place, buy/beg/borrow/steal ventilators and PPE, and build testing and contact tracing capacity.

It's far too widespread right now to successfully control via contact tracing and isolation, even if we had infinite testing capacity. The optimal use of testing right now is whatever produces the largest number of positive results, so that large case counts stay in the headlines, helping to ensure political support for stay at home orders. And that's more or less how the tests are being allocated.

My firm (options market making, not exactly HFT, but close) has most people working from home successfully, but some are having problems with flaky internet. If I was the owner of a big shop I would probably want at least a skeleton crew in the office. A regional Comcast outage could easily take 90% of the remote employees offline.

I think you are correct in your basic assertion that no one wants manual memory management for its own sake. What they really want is sufficient performance for their use case. The benchmarks you usually see are throughput oriented, and on small heaps. If you have tight latency budgets and/or huge heaps, the performance is not close.

Optional manual memory management sounds great, but I'm skeptical it would work well in practice. The reason is that if the language default is GC, libraries won't be designed for manual memory management, meaning it will be hard for your manual code to interact with data structures created by non-manual parts.

Love the concept. If the author is hanging around, are you working on this full time, or is this a side project?

Despite being legal under local law in a number of states, possession of even small quantities of marijuana remains illegal under federal law[0].

In the words of the National Security Adjudicative Guidelines[1]: "The illegal use of controlled substances... can raise questions about an individual's reliability and trustworthiness, both because such behavior may lead to physical or psychological impairment and because it raises questions about a person's ability or willingness to comply with laws, rules, and regulations."

Whether or not you agree with the policy, it's still the policy, and if it's enforced when non-famous people with security clearances use drugs (it is), it should be enforced when famous people with security clearances use drugs.

[0] 21 USC § 844: https://www.law.cornell.edu/uscode/text/21/844

[1] the official policy defining the criteria by which security clearances are granted: https://fas.org/sgp/othergov/intel/sead-4.pdf

This particular article basically uses Edge Computing as a synonym for client-side computing. I've more commonly seen the phrase Edge Computing used in a different and more interesting way. Basically, a typical web app architecture would have servers for dynamic content in 2 or 3 places, mainly for redundancy, and use a CDN to distribute static content. The CDN in this case is the Edge, but there's no compute: just caching of static assets. That's not a great fit for some applications (those with high bandwidth or low latency requirements on dynamic data especially), but there's not really been provider support for that, so if you wanted to do it you had to deal with your own hardware in a large number of locations. If a cloud provider adds Compute to the Edge (ie Lambda@Edge from AWS, Cloudflare Workers), this architecture becomes doable for small shops.

Yeah, I think the Microsoft messaging would naturally lead you to that conclusion, though generally they avoid saying it explicitly. I think in a year WPF and WinForms will probably work great under .NET Core, but I wouldn't use either for real work right now. My recommendation would be to use .NET Framework for GUI work right now, but do it in such a way that the upgrade path is easy. The way to do that is to create the project under .NET Core, and then hand edit the project file to change the <TargetFramework> element from "netcoreapp3.0" to "net472". That will give you the new project file format and make it easy to upgrade when they finish getting the bugs out. The other main thing you should do is to create any libraries as .NET Standard 2.0 libraries, which work with both .NET Framework and .NET Core.

The result here is from a survey of Mensa members. I'm skeptical Mensa members are representative of the high IQ population as a whole.

If you were trying WinForms, the designer does not yet work for .NET Core apps. The WPF designer is supposed to work with .NET Core, but there was a bug in it in the first VS release that might have been the cause of your issue[0].

I also had problems doing anything other than toy WPF projects. IMO the .NET Core support for the GUI frameworks is not ready for prime time. Building GUI apps in .NET Framework is still a great experience though.

[0] https://developercommunity.visualstudio.com/content/problem/...

It's not actually permanent. Your bank almost certainly has the right to change it daily, and once their asset acquisition goals are met, they probably will.

I don't think it's actually true. See my response to GP.

> Now, if the risk-free rate is negative, you might reasonably ask: Why doesn't the bank just keep cash? And the answer to that is that they're legally prevented from doing so.

Some cursory research suggests this is not the case. The Federal Reserve website states: "Depository institutions must hold reserves in the form of vault cash or deposits with Federal Reserve Banks." [0]

[0] https://www.federalreserve.gov/monetarypolicy/reservereq.htm

If you want to know, it's probably in the prospectus. I'd be shocked if termination called for anything other than liquidating the holdings of the trust and distributing the proceeds pro-rata to shareholders. Even if something weird was supposed to happen, there would be at least 20 years of notice for everyone to get their money out.

The reducing missed cases of AKI from 12% to 3% stat was a result from the introduction of the Streams app into the AKI workflow. The model used was an existing NHS AKI model, not the new model discussed in the first part of the article.

Not only does submitting a claim prevent you from suing Equifax, the only way to preserve your right to sue is to actively exclude yourself from the settlement. See FAQs 19 and 23.

19: https://www.equifaxbreachsettlement.com/faq#q-19

23: https://www.equifaxbreachsettlement.com/faq#q-23

Way late reply you'll probably never see, but I'm fairly sure it's possible with the U2F protocol and existing hardware. I simplified the algo so as not to obscure the big picture, but I have actually worked out all the details and identified hardware I think would work. (I seriously considered trying to build a business around the idea and worked through the feasibility in a lot of detail. In the end I decided it was completely feasible but a crapload of work and limited marketability.)

You've correctly identified most of the tricky bits: the real algo involves key agreement plus symmetric crypto, it's a tight fit size-wise, and not all hardware has the necessary capabilities. However I think you're wrong about the conclusion. Can't say for sure though, as I didn't produce a working prototype. If you actually see this and want to discuss it in more detail you can find my email in my profile.

Correct, in any specific case. But Apple's playing a repeated game, and any given customer is playing a one-time game. Apple can reap the deterrence benefit in future cases, but the customer doesn't really get any reputational benefit from playing hardball, and so has to evaluate the wisdom of suing or not suing based on the outcome of just the single case at hand.

In any one specific case, fixing a litigious customer's laptop is definitely cheaper for Apple. But by having a policy of fighting legal claims vigorously regardless of merit, they can make it wildly uneconomic for customers to sue, which decreases the probability of getting sued. I suspect the economics of their behavior work quite well overall.

A rollback without understanding is definitely risky. An uninformed rollback is one of the factors that killed Knight Capital Group in 2012. For those not familiar, the actual problem was they failed to update one of a cluster of eight servers, and the server on the old version was making bad trades. They attempted to mitigate with a rollback, which made all eight servers start to make bad trades. In the end they lost $460 million over the course of about 45 minutes.

The full report is here if you're curious: https://www.sec.gov/litigation/admin/2013/34-70694.pdf