HNHacker News
TopNewBestAskShowJobs

peterwwillis

1 karma · joined March 4, 2021

submissionscomments
peterwwillis··on Spy agency ducks questions about 'back doors' in tech products
If, hypothetically, you worked for one of these vendors, then, hypothetically, you might find that while waiting for a shipment of new product test gear to your lab, a shipment from California to Maryland may get waylayed - by U.S. Customs, in Texas. Two weeks later you finally get your gear. And you don't talk about it outside your immediate team. Such is life in a big company.

Hypothetically.

peterwwillis··on A Theory of Software Architecture
Some designs don't require your own code to pass around state between functions, because each function can handle state in its own way without dependence on other functions. I've written a lot of such simple scripts where I just need to execute a series of tasks that aren't necessarily related to one another, but do follow each other.
peterwwillis··on A Theory of Software Architecture
I don't really fully understand how the kernel scheduler works, but I know in general how it works, and so I can write software using it. In this sense, a simple implicit abstraction is easier for me to understand than one where I'm peeking under the covers, so to speak.

Even just seeing a few variables or objects passed around, I still don't know exactly what it's doing, or hiding. I can more quickly understand the high level idea with the former example without the details potentially confusing me.

peterwwillis··on John Gray: 'What can we learn from cats? Don't live in an imagined future'
It always struck me as very weird that Buddhists want to escape suffering. That's just saying you want to escape life. Suffering is beautiful and one of the best parts of life (the only thing better is momentarily overcoming it). But having neither is just sitting around waiting to die.
peterwwillis··on John Gray: 'What can we learn from cats? Don't live in an imagined future'
Also, man is a laughing stock to the ape. "What are you doing without hair? Why so weak? Ugh, what is that you're eating?!" I can scarcely imagine what kind of unsustainable sacrifices the Übermensch will make in service to "progress".
peterwwillis··on When will web browsers be complete?
The browser becomes the OS and then that OS will get a browser.

Turtles.

peterwwillis··on A Theory of Software Architecture
The former example was much easier for me to understand than the latter, even if it wasn't factored well. But I'm assuming that the functions aren't mucking about with global state or something equally distasteful.
peterwwillis··on Restrict Access to your internal websites on AWS with BeyondCorp
I haven't implemented this extensively, but there are sometimes problems with setting this up in a corporate environment where there's a need to do things in a uniform way.

- Often if you use a central database such as Active Directory for your internal users, you may need to set up your OAuth endpoint (say, an AWS Cognito User Pool) and then have an AD admin allow that user pool to be a relying party. This means there is some lag time to setting this up, and that it isn't done with automation. So if you have an application you want to spin up a custom domain with a temporary user pool, test something, and then destroy it later, it's probably not going to work without some custom workarounds. No need for this with a traditional VPN.

- If you have 50 different apps, with 50 different URLs, you're going to need to do the above 50 times. Also have 50 staging and dev portals? Same deal. Now try managing changes to all of those at once. The manual steps, plus all the integration with all the product teams, now means this whole thing is becoming burdensome. This is a lot more work than just "use this VPN client and whitelist these CIDRs".

- If you're working for one of those weird companies that loves to do split-horizon DNS with lots of custom internal-only domains, guess what? Probably not gonna work without a VPN.

- Onboarding can be complicated. You now need to help your users manage their accounts, such as password resets, being a part of different domains, using a supported device, MFA registration, troubleshooting internet connections, using a supported CLI tool for non-web-interface APIs, etc. Versus just saying "are you connected to the VPN?".

- A VPN allows a central network team to manage access control across the entire network (or wherever that network team manages networks). BeyondCorp needs to be managed for all of your products by one team, or you may end up with an uneven and difficult process of supporting users across teams. A lot of companies (maybe most) are just not set up to allow independent product teams to manage internal user access. Even then, authorization maybe, not probably not both AuthN + Z.

- If you have a single domain serving multiple apps in multiple URLs requiring different authorization, things can get more complex.

Ultimately I think most of the protocols used for BeyondCorp today have too many drawbacks to say we could all drop our VPNs in support of it. We'll probably need at least another generation of protocols and management workflows in order for it to become the new norm.

peterwwillis··on A Kubernetes operator to sync secrets from AWS Secrets Manager
My negativity comes from this specific solution. My two main problems with it:

1. It's the exception, not the rule. Most of the time there aren't collaborative solutions. Most of the time people work in their silos and make something work for themselves. Often it's corporate-funded, and those corporations have no consideration for others being able to pick up the work once they've abandoned it. The fact that there's 10 of the same thing to integrate is proof enough. It's just gotten so ridiculous that somebody finally had to address it. Turns out it's a company that has to deal with all of those implementations anyway.

2. This is one tool figuring out how a bunch of other tools can integrate into it. It's like what I'm proposing, if you assume the least work possible to achieve maximum laziness for your own specific tool and use case.

Their solution wasn't "hey, it looks like a lot of things use secrets. maybe it'd be cool if we made one way for any system with secrets to interoperate with any other system, and try to get it adopted by existing vendors." Their solution was "hey, we just need to get all these other things to work for the one tool we're using." Rather than "how can we make this more composable, more compatible, easier to implement, for everybody... even if they're not using our tool...", it became "how can we make this easier for just us?"

That's what is always happening, has always been happening, in this space, for nearly 10 years. Either everybody latches their custom tech onto a single platform and no real open source solutions get made, or everybody spits out incompatible, overbuilt, underthought, opinionated solutions for their own problem. We don't build standard solutions in the cloud world, we build log cabins.

I mean, just read the PR. They're talking about coding into this 'standard' support for each different implementation. Like "this is a vault secret, and this is a gsm secret." That's the opposite of what I want. I want it to say, "this is what a secret looks like. now you figure out how to use it.", or, "give me your secret. I don't care who or what you are, because we all speak the same secret language." That is what an internet standard is supposed to be. Not "this is a bind9 DNS record, this is an AD DNS record, this is a RedHat DNS record, this is a Route53 DNS record".

The "container landscape" shouldn't remotely resemble what it does today. The idea of a "container" should have been standardized in an agnostic way, without requiring all the (admittedly useful, but also completely unnecessary, and often burdensome) features Docker threw into their tool. Yes, many people's lives were made easier. But a whole lot of other lives were made harder, to the point of small economies built out of badly re-implementing and custom-integrating into one precious and incompatible concept.

I would call Kubernetes the open source Active Directory, but Active Directory is literally ten times more standards compliant.

peterwwillis··on A Kubernetes operator to sync secrets from AWS Secrets Manager
This is why we need standard protocols and data formats for these common systems. If there were just a "secrets protocol" or "secrets data format", any program could just implement it for input and output, along with a standard interface to perform the actions (a single URL for example). It used to be commonplace to just write an RFC for what you were doing and then other people would use the RFC. Which wasn't perfect, but nothing ever is...

Instead, the standard solution today is custom integration, which leads to a lot of reinventing the wheel and incompatibility with extremely similar products.

peterwwillis··on Aim for Operability, Not SRE as a Cult
I don't think anyone argues that SRE is wrong, but I think most people will be wrong if they assume most places can or will implement it. It's just like DevOps in the sense that it's a corporate efficiency strategy, but it's never seen as all that necessary. Execs just see it as another Six Sigma, something to banter about at the golf club. And nobody outside of sysadmins and managers really get it at all.

I think the more companies rely on technology, the more screwed they are, because they don't realize how much harder it is to run a business dependent on technology. The fact that SRE exists proves that, I think. The system is by default a garbage fire, because by default you can't push back on product in order to fix your tech debt. Most other businesses seem to grasp the concept that they need to perform maintenance on their machines or they won't be in business for long, but as soon as it's "the magical box with the 0s and 1s inside", they pretend it doesn't require long term investment or a different strategy.

peterwwillis··on The businesses where Google is biggest and the ones where it isn’t
> the only viable option

Let's not lie to ourselves. It's the option you prefer, not the only viable one.

It's also a lie to pretend that Google is any more evil than DDG. Both are capitalist corporations marketing a story to convince people to give them money (in this case, to convince you to make money for them). Neither of these organizations have morals, they have profit margins and consumer strategies. "Privacy" is just the tech equivalent of "Organic", a marketing tool to convince you to choose their brand. Eventually Google will get on the "privacy train" and people will completely forget they were "evil", just like with Microsoft. It's all so much Eurasia versus Eastasia.

peterwwillis··on Many of the root certificates on Windows are not needed
No, to exploit this, you just need to either 1) work inside a CA, 2) be a spy and infiltrate the CA, 3) use your nation-state powers to force a CA in your borders to do what you want, or 4) be a clever hacker hack a CA. There are "only" 330 different complex organizations in probably a hundred-plus countries to try, so I guess this is pretty unlikely.

But even if it is, you can actually hack the entire system with just a BGP hack. Today, if you issue some false BGP routes, you can MITM LetsEncrypt (or one of the other 329 CAs) to sign a valid cert for you. It's not hard because validation is automated and depends on proving you own IP space... and IP space is just what your router's BGP table says it is at the moment you issue a cert.

I would not depend on the kindness of banks giving back money on very large scales. They care about themselves more than they care about you.

You're right, DNS registrar and nameserver attack is another great attack and usually they're pretty vulnerable, so that's another good option.

peterwwillis··on Many of the root certificates on Windows are not needed
> Isn't there a better way?

There's always a better way. But when you're going up against gigantic incumbents with way more power than everyone else, you don't get the better way. You get the way the powerful decide it will work. It's like politics, or economics. When a system is so large and unwieldy, significant change can't come unless the ridiculously powerful minority forces it. If they have no reason to do so...

peterwwillis··on Many of the root certificates on Windows are not needed
Yes, it's called SaaS, and their "ad-blogs" make up about 20% of all the top stories on HN. Startups that charge you to use their proprietary software but write good clickbait or Show HNs. (I'm not hating, I'm just saying this is not surprising in any way)
peterwwillis··on IKEA to buy back used furniture in recycling push
I used to buy a lot of Ikea furniture (and take a lot of free Ikea furniture that ends up on curbs in big cities around moving season).

Then I learned how much cheaper, modular, and more robust your furniture can be if you invest in some hand tools and built it yourself. You have a higher initial outlay, but the results last longer, can be repaired much easier, as well as broken down and repurposed. Most furniture can be assembled with a cheap hand saw and chisels, or if you're lazy, skilsaw, drill, drill bit set, and screws.

I don't know that i'll ever buy a piece of furniture again unless it's second-hand and deeply discounted. There's so much good shit out there already.

peterwwillis··on No-till no-herbicide farming system in trial since 1981
I've gotten massive amounts of food out of small raised beds with nothing more than compost tea and "plant food". Is it really not possible to consistently produce good yields from tightly packed beds with crop rotation on a 1 acre plot [without pesticides/herbicides]? I know it wouldn't scale up with typical large-scale farming monocultures, but for a mixed family plot?

Aside: it seems we've long been able to feed the world (in terms of meeting caloric and nutritional needs) based on American farm capacity alone, it's just that nobody wants to pay for it.

peterwwillis··on This page is a truly naked, brutalist HTML quine
It can receive input and create output. The browser is the interpreter for scripted HTML programs.
peterwwillis··on Emerging Architectures for Modern Data Infrastructure
Sounds like you need 1,000 nodes to do 1Bpps without edge computing. With some compression at the edge, it'd be closer to 150-250. The limits of a conventionally architected network make it more annoying than it needs to be.
peterwwillis··on Thoughts on Conway's Law and the Software Stack (2019)
Communication problems usually come down to people not understanding each other, which has many causes of course. Most of these have to do with people being all different. When people are all the same, less communication problems.

But having people all the same then introduces non-communication problems. Lack of genetic diversity is a potential catastrophe. Lack of outside perspective makes novel thinking difficult. Lack of change breeds complacency.

So you need your software to work pretty much all the same way, but you need diversity in how it works to survive environmental challenges.

Nature has a solution for this (as in most things): evolution. You mutate a bit every so often and eventually something works better. Then you stick with that and repeat. It's a continuous improvement cycle driven by measurably better outcomes from experiments.

This can work for hardware too but it's definitely not as straightforward. But no matter what you use it on, you can't have some clunky, hierarchical, reactive, conservative, single-minded power structure holding the process back. The stack will always suffer as long as there's a giant anchor at the top keeping it from being agile and mutating. It's not so much a cultural thing as an organism; if the organism can't agree on how to grow, it's not gonna fare well against competition.

peterwwillis··on Canonical introduces high-availability Micro-Kubernetes
> I’ve never seen a production Kube system with HA etcd go down due to non-human error

I have, with OpenShift, several times. The worst one took hours to get it partially back online, days to fully recover. At least I got to have one free drink before I had to leave our holiday party to start fixing it. (Ok, technically it was non-prod; luckily prod was ECS... long story)

> I don’t believe single instance is going to give you better availability when single machine faults happen

Depends on your definition and context of availability. It is possibly the simplest thing to destroy any existing nodes (and in the event that a node can't be destroyed, null-route it and/or disable its network port) and bring up a new node. So single instance is fairly easy to recover. But with multiple nodes, for each function of each instance that is required to reach consensus, the likelihood of consensus failure increases; you actually need more nodes and variety in the cluster to resist consensus failure. Yet ironically, the more nodes you have, the higher the probability of failure. In the end, the real-world reliability of the system is based on additional factors besides the network model.

peterwwillis··on AWS forked my project and launched it as its own service
Twitter clap threads clap make clap bad clap articles

In a month / year / etc when this thread gets deleted, there'll just be a HN comments section speculating about a missing twitter thread. The least we could do here is ask posters to capture the thread in a blog post along with a summary of points, so it doesn't have to be re-created in the comments here. This is potentially an important post that we want a historical record of, and Twitter is just not built for that.

peterwwillis··on Canonical introduces high-availability Micro-Kubernetes
You don't need distributed consensus to have a cluster of reliable services. We've had large-scale clusters of applications for what, 30 years? And distributed consensus being generally used by a few players for about 10 of those. Etcd is neat, but in no way mandatory for what it was even invented for. 99% of people just don't need it.
peterwwillis··on Why Everyone Builds Internal Kubernetes Platforms
mostly it's Conway's Law
peterwwillis··on Dockerfile Security Best Practices
This is (imho) a huge flaw in the concept of a "container". I don't think most people comprehend how much crap is going on in the background.

For most container purposes, host networking and the default process namespace is absolutely fine, and reduces a lot of problems with interacting with containerized apps. 95% of the use case of containers is effectively just a chroot wrapper. If you need more features, this should be optional. This would also make rootless federated containerized apps just work. But nobody wants to go back to incremental features if Docker gives them everything at once.

peterwwillis··on 2.1M of the oldest Usenet posts are now online for anyone to read
One of the things it eventually became was the world's largest, fastest and most reliable media-pirating service, paid for and hosted by your ISP. Eventually the binary groups got dropped by ISPs, but you could pay to use someone else's servers and pull a couple hundred gigabytes of content at fiber speeds. This all while people were struggling for weeks to get a single torrent to finish. Of course, FXP servers and IRC D/C were the OG pirating method, but usenet binaries were accessible to pretty much anybody and in no way illegal ("I'm just downloading the news!"). Though it also seemed like nobody even noticed anything illegal was going on, because most people were stuck on P2P alternatives even though they sucked in comparison. I'm pretty sure Napster succeeded because normal people couldn't comprehend Usenet.
peterwwillis··on Keeping CALM: When Distributed Consistency Is Easy
Another way to think of a distributed system is as a threaded application on a multi-core CPU. Break down the interactions within the application by all the logical boundaries and you have a basic distributed system. Typically the input and output are via a single controlled mechanism and state is handled in a single place in memory during the series of events for a particular operation, so consistency is just assumed. But give all the series of events their own state, and each event a different input and output in random ordering, and you have distributed system problems. Threaded apps have similar problems to solve and have their own mechanisms to deal with these problems, so you can compare them and see how multiple mechanisms may be needed to compute safely.
peterwwillis··on Fuck You Sunday
> Capitalism is a system that runs on human misery,

The last two capitalistic organizations I've worked for, all my co-workers were cheery, friendly, helpful, and enjoyed their jobs. The people frustrated were the ones that did their jobs as good as they could and were frustrated by a lack of excellence in the face of large scale and complexity. I wouldn't call this misery. And one of those companies was literally rated the most hated corporation in America. I've also worked for places were people were lazy, bureaucratic, and did the least work possible to collect their paycheck (heck I was that person). Ultimately corporations are made up of people, and there's all different kinds of people out there, so there's also all different kinds of corporations.

> led by a class of people who are completely unaccountable to anyone but themselves, who therefore have no incentive to make things decent for the workers.

They're accountable to shareholders and government regulatory bodies, and in very rare circumstances, to unions. Certainly there should be more accountability within the organization from the bottom-up, but it's fallacious to think they don't answer to anybody. The actual incentives vary, but usually stem from either competition, or seeking to drive quality and efficiency. Business management is more art than science.

peterwwillis··on The Serverless Revolution Has Stalled
Serverless is the most overhyped cron job runner in history.
peterwwillis··on We spent two weeks hunting an NFS bug in the Linux kernel (2018)
You were unfairly downvoted. This is an important question.

Sometimes NFS is your only option. When you have legacy applications which nobody is going to rewrite, and you need to scale some operations on a set of files over many nodes, NFS works. Its failure nodes are well documented, so you can plan for and work around them.

It's also supported by most vendors as the default networked filesystem. Whether you're using a desktop PC, a million dollar NAS, or cloud storage, it supports NFS.

It's also acceptable when time is of the essence and you just need to get something working now and don't have time to build something better. I rather like "good enough solutions" as a way to get teams "shipping", with an agreed-upon plan to replace it later.

But as someone who's used it for nearly 20 years, internally I start screaming and throwing things whenever someone suggests using it. Externally I propose redesigning for a more efficient solution :)

← PreviousPage 3 of 34Next →