HNHacker News
TopNewBestAskShowJobs

charleslmunger

1,327 karma · joined March 3, 2013

submissionscomments
charleslmunger··on How I write HTTP services in Go after 13 years
My experience has been the opposite - we've been maintaining large (tens of millions loc, thousands of developers) mobile app codebases with this approach for about 8 years now, and results have been much better than what was done before.

A number of bad decisions were made and later discovered over time, and it was much easier to address them with automatic dependency injection available than without. Using as "big" of a test as you can without regressing speed or flakiness gives you really solid, non-fragile tests, which makes large changes tractable to do safely.

I don't know whether certain poor designs would have been spotted sooner, but this certainly made the code easier to change and resulted in more useful and lower maintenance tests.

charleslmunger··on How I write HTTP services in Go after 13 years
Maybe this is heresy in the Go community, but this is a problem that automatic dependency injection solves.

https://dagger.dev/hilt/testing-philosophy.html

The biggest problem with the "pass nil for unused dependencies" approach is that when you modify some code to actually use one of those dependencies when it didn't before, you have to go back through every test and populate it.

charleslmunger··on Huawei prepares to split from Android on consumer devices with HarmonyOS Next
https://consumer.huawei.com/en/opensource/detail/?siteCode=w...
charleslmunger··on So you think you know C? (2016)
When would you want a loop that does not terminate and does not perform I/O, read/write a volatile, synchronize with another thread, or perform an atomic operation?
charleslmunger··on We switched to Java 21 virtual threads and got a deadlock in TPC-C for Postgres
"Why not to use java virtual threads" -> Fundamental features of the language treated as implementation details in libraries for more than two decades cause deadlock.
charleslmunger··on 32GB of RAM is becoming the standard
>computers with little RAM have to swap this data to the hard drive

The hard drive? What machine is sold with a hard drive as swap in 2024?

charleslmunger··on Ask HN: Stay at Big Tech or leave?
Consider switching teams. At every big tech company there are different teams that spend more time on different tasks. Keep your compensation, change your work. Implement new features in a kernel, or new optimization passes in a compiler. Build a UI for an new feature.

But if that doesn't work, you should leave. It's is a tremendous opportunity cost to spend your early career not learning new things.

Jonathan Blow has strong opinions[1] on this.

[1] https://m.youtube.com/watch?v=nL8GWU9M8LY

charleslmunger··on Amazon receives FDA warning letter for supplements with undeclared ingredients
As for that "100% safe" homeopathic stuff... https://www.fda.gov/news-events/press-announcements/fda-conf...
charleslmunger··on Ask HN: Why do major operating system installs default to HTTP instead of HTTPS
All the packages are signed. But one practical reason might be that a computer doing a fresh install may not have an accurate system time, which TLS depends on.
charleslmunger··on Are You Smarter Than An LLM? (Quiz based on the most popular LLM benchmark)
This was very cool until I realized that a significant fraction of the questions have incorrect answers. Two for two biology related questions were wrong.
charleslmunger··on Hardware Intrinsics in .NET 8
This is not always true - for many GCs, threads need to suspend at least some of the time. If the compiler can't prove a small bound on the iterations of a loop, it needs to insert safe point instructions.

Also depending on the GC, read or write barriers may be required.

Lastly there are circumstances in generational moving GCs where there is extra overhead due to checking to and from space - for example if you attempt a compare-and-set on a reference type, you need extra code to check if your CAS failed due to comparing the from-space and to-space addresses of the same object.

I work primarily in garbage collected languages, and there are many benefits - safety, throughput, easy ABA avoidance in lock-free code - but using a GC carries tradeoffs.

charleslmunger··on Asus Intros GeForce RTX 4060 Ti Video Card with Integrated M.2 SSD Slot
This actually would be very cool when combined with an m.2 NIC [1] - ITX boards only have one PCIE slot and their built-in m.2 slots rarely have clearance for that height of heatsink, so getting both a GPU and 10gb networking is hard.

[1] https://www.aliexpress.us/item/3256805824691996.html?spm=a2g...

charleslmunger··on Designing a new concurrent data structure
DirectByteBuffer would allow direct pointer access to the backing memory. When used with allocateDirect, you can rely on Java's GC to manage the memory. Not sure if it's a perfect fitnfor the use case but it's simpler to reason about.
charleslmunger··on Matt Levine: Ripple Is a Security and It Isn’t
>When loan contracts are enforced transparently and in accordance with terms that are immutably encoded, there is virtually zero risk of fraud and mistakes.

Like solend?

https://app.realms.today/dao/7sf3tcWm58vhtkJMwuw2P3T6UBX7UE5...

charleslmunger··on Matt Levine: Ripple Is a Security and It Isn’t
>20% inflationary APY with an algorithmic "stable"coin

Was a reference to terra and the anchor protocol. And while it's true that its flaws were less obvious than some truly stupid tokens from 2016, I don't think that's a fair characterization of the problem. In some ways it was actually a more obvious failure than dumb 2016 tokens. If I announce a perpetual motion machine, you don't need to actually analyze its schematic to know that it can't work the way I say. The same is true of a self balancing token pair scheme. A high school economics class is all that's required to understand why it can't work.

An overcollateralized loan is traditional lending. It's a software-implemented mortgage. How reliable it is is dependent on the risk profile of the backing assets and the creditworthiness of the borrower. A liquid market is a workable but imperfect way to assess value. It cannot measure risk - the information simply is not there. Volatility is not risk.

As for what portion was baseless... I think that's a fundamentally subjective question. From my perspective, the vast, vast majority in both years was baseless. I look at the entire cryptocurrency space as a place where interesting tech could have found value, but it's so infested with fraud and speculation that it can't.

charleslmunger··on Matt Levine: Ripple Is a Security and It Isn’t
Terra had an ICO in 2019. SafeMoon had its pump and dump in 2021. NFTs were sold at auction directly to the public. 3AC is the existence proof that token investors were not all sophisticated.

>CEL [...] wasn't an obvious scam at the outset

From the Celsius website:

>Earn up to 17% APY and get paid weekly when you HODL your coins with Celsius

It is not possible to pay 17% on deposits in a time of historically low interest rates while taking the low level of risk Celsius was claiming. The specific mechanism that was claimed, namely that they were taking a lower spread than banks, was obviously absurd. On the same website there is a diagram that is literally a circle explaining how CEL returns work.

>reputable and established exchanges like Coinbase

Gemini is supposed to be reputable and established. How are investors in its Earn offering doing now?

My point was that the total number of people who gave away sovereign currency in exchange for cryptocurrency they did not intend to use directly (i.e., buying a pizza, publishing smart contracts, etc) was much larger in 2021 than 2018. I think we probably have different ideas about whether speculating in a manipulated market rife with fraud counts as getting scammed, but surely we can agree that crypto had much wider reach in 2021.

charleslmunger··on Matt Levine: Ripple Is a Security and It Isn’t
CEL is a token,FTT is a token, SafeMoon is a token, the T in NFT is for token. 3AC didn't list a token but they were an investor in obviously stupid token schemes - 20% inflationary APY with an algorithmic "stable"coin. The facts are that a lot of money was poured either directly into stupid/fraudulent tokens or into enterprises that made a big deal out of issuing their own stupid/fraudulent tokens. "only reputable teams were able to raise money" is simply not true.

>the situation was much worse in 2010-2016

That might be true, but the people being scammed with lambo dreams in 2016 were a fraction of those getting scammed in 2021.

charleslmunger··on Matt Levine: Ripple Is a Security and It Isn’t
>By the end of the brief experiment, which ended with the SEC's entry into the sector in early 2018, the typical token investor was highly discerning, with the result being that only reputable teams were able to raise money.

* FTX * CEL * 3AC * SafeMoon * The entire NFT "ecosystem"

charleslmunger··on Shortening the Let's Encrypt chain of trust
What extensions other than NEON are you thinking of? Neon was made a CDD requirement for Android 6.0, released on 2015. Even before that it was almost ubiquitous. The last shipping devices in any volume missing NEON were Tegra 2 tablets from 2011.
charleslmunger··on Repair adventure: A single white LED causing stuck keys on a RN988 keyboard
The capacitor plague was real. I had a 22" monitor that died after four years. Some enterprising person had set up a webshop where you could put in your model number and they'd send you a little bag of capacitors that matched. I paid $15 for what was probably a few cents of capacitors and that monitor lived on for many years, despite my sloppy soldering.

There were a few that were visiting bulging or had leaked, but I replaced all of them just in case.

charleslmunger··on Android application development, how is your world?
Like anything it depends what you're building. If you're just making an app that's basically a WebView, or a game using a cross platform engine, you will not find much that is specifically creative about Android development.

Android offers a huge and open api surface, and you can replace many user facing parts of the OS - alternative dialers, launchers, etc are fair game. There are fewer bounds on your creativity than many other platforms.

Android also offers an enormous variety of hardware, which includes very low spec affordable devices. Getting good performance on these devices requires effort put into efficient software design and an optimized implementation, and that constraint makes programming more enjoyable for certain types of people.

I've been developing for Android for about 11 years. Like all platforms there are bugs, there's legacy code, not every task you do is equally intellectually rewarding, but I've stuck with it because it's fun and I'm always learning new things.

charleslmunger··on FreeBSD spends 7% of its boot time running a bubblesort on its SYSINITs
I submitted some posts to this but it doesn't seem like the author has updated it. I probably run into one of these once a month!
charleslmunger··on iOS 17 app sideloading might only be available in Europe
It can be simultaneously true that a warning can be a barrier to adoption of your software product versus a competitor, and that a warning is not an effective barrier for a user who thinks they're installing an unreleased video game or going to receive millions of dollars of crypto from foreign royalty.
charleslmunger··on Displaying my washing machine's remaining time with curl, jq, and pizauth
This is absolutely the way to do it for washers and non-electric dryers.

I configured my automation to notify me when both the washer and dryer are idle.

charleslmunger··on An EEVDF CPU Scheduler for Linux
Android sets the priority of the top app's main thread and render thread to -10. Devices often also configure a cpuset reservation for the top app.
charleslmunger··on Recursive (Re-Entrant) Locks (2013)
The @Contended annotation will pad the memory layout of fields to fill a cache line and avoid false sharing.
charleslmunger··on Magic beans, Bahamian penthouses, old-fashioned fraud and other important SBF-in
>In the event that whoever runs Tether has committed financial fraud on a large enough scale to lose, say, 5% if the total value,

Why is this phrased like a hypothetical?

>In fact Tether reserves were not “fully-backed” the majority of the time.

https://www.cftc.gov/PressRoom/PressReleases/8450-21

charleslmunger··on Golang disables Nagle's Algorithm by default
Clients should not be nagling unless the connection is emitting tiny bytes at high frequency. But that's a very odd thing to do, and in most/all cases there's some reasonable buffering occuring higher up in the stack that the nagle's algorithm will only add overhead to. Making things worse are tcp-within-tcp things like http/2.

Nagle's algorithm works great for things like telnet but should not be applied as a default to general purpose networking.

charleslmunger··on Eufy’s “No clouds” cameras upload facial thumbnails to AWS
At least on android, the payload can be encrypted - the notification that gets shown is built by the app.
charleslmunger··on Tell HN: Apple refuses to delete a personal domain Apple ID that I didn't create
I had this same experience several years ago - someone signed up for an apple ID using my corporate email. I escalated it via the phone tree, and eventually got a phone call back from some sort of manager who was able to detach my email without otherwise disrupting the person's account.

I suspect that your best bet is to use the phone support via TTY. As a non-apple customer, I was very impressed by their dedication to resolving my issue, and I hope/suspect that if you can get your case escalated they will find a way to communicate with you without audio or sign. But getting to that point may require an initial interaction with the phone system. My experience did not lead to any linkage between my phone number and the problematic iCloud account.

← PreviousPage 6 of 16Next →