HNHacker News
TopNewBestAskShowJobs

vilunov

310 karma · joined August 1, 2019

submissionscomments
vilunov··on Maybe Rust isn’t a good tool for massively concurrent, userspace software
"you just have to know what to do" is the problem. You can call any color from any color, but for some colors it's trivial, e.g. sync function from a sync or async one, or a non-failing function from a failing or non-failing one.

I don't want to be able to call fallible function from an infallible one trivially, I want the compiler to force me to specify what exactly I'm going to do with an error if it happens. Likewise for async-from-sync: there are many ways I could call these: I can either create a single threaded executor and use it to complete the future to completion, or maybe I want to create a multithreaded executor, or maybe I expect the future complete in a single poll and never suspend and I don't even need a scheduler.

vilunov··on Maybe Rust isn’t a good tool for massively concurrent, userspace software
I disagree with you in the last point, async is definitely painful for end users. It indeed feels like you're using a completely different language, which has Rust's core features removed – lifetimes and explicit types, sprinkled with a mess of Pins on top.

You cannot run scoped fibers, forcing you to "Arc shit up", Pins are unusable without unsafe, and a tiniest change in an async-function could make the future !Send across the entire codebase.

vilunov··on Maybe Rust isn’t a good tool for massively concurrent, userspace software
I want colored functions. I want to know which code is running synchronously and which doesn't, which raises errors and which doesn't. Color is just a description of the function's properties (and effects) and how it's compatible with other colors.

There is also nothing fundamentally bad with cooperative scheduling in scope of a single process.

vilunov··on NSO group iPhone zero-click, zero-day exploit captured in the wild
What exactly is "an exceptionally restrictive sandbox"?

There are virtual machines such as JVM, V8, or even QEMU. These are sandboxes, which run either some special bytecode or native code with extreme performance drawbacks. Media decoders are performance- and energy-sensitive pieces of software in the end.

And media decoders actually ARE sandboxes of sorts. They are designed to interpret media formats, sometimes even Turing-complete bytecode in retrictive and isolated environments. And like any sandboxes, they too have bugs.

vilunov··on ForgeFed
I don't think that it's up to a person or even to a group of people to decide whether one protocol is strictly better than a different one. Only by providing a decent competitive environments can we really find out which one is better.

Federation (and fediverse) can run on different protocols, by deciding to implement ActivityPub only we are still living in a walled garden of sorts, only this time the bubble is not bounded by a single service provider, but by a protocol and its sphere of reach.

AP is certainly all the rage now, but email-based federation is older and it is already implemented in sourcehut. Do other software forges implement multiprotocol federation? Would they agree to implement email-based federation?

vilunov··on ForgeFed
https://drewdevault.com/2018/07/23/Git-is-already-distribute...

I understand his reasoning, but there needs not to be a single protocol for federalization. You can support both email and AP and let users choose what is best for them. "Email simply being a better choice" is a controversial opinion (as most of other statements there). Protocols can evolve and compete with each other, but only if software gives them space to do that.

vilunov··on ForgeFed
I host a number of services for me and close friends and family, and I limit the access by manually creating accounts for them. I could enable OAuth via 3rd party service, but that would mean I'll have to pay much more attention to setup of ACLs, as these users will have a new account created for them with a link to their Github account.

Besides, I want to stay more or less independent from 3rd party services, particularly I'm bothered by the need to explicitly register an OAuth client. AP has a much better flow in that regard, services can interoperate without a given permission to do so.

vilunov··on ForgeFed
Before I learned about that I usually dismissed Forgejo as an unimportant fork which will die off pretty quickly, but support of ForgeFed makes me seriously consider migrating from Gitea. I still believe that Gitea's org strategy is the right one in the long run, however.
vilunov··on ForgeFed
I'm looking forward to this. As of now, if you want to contain development of your software to your personal gitea instance, you are actually raising the complexity for new contributors, as they have to create a new account on your instance to create issues and pull requests. This is even worse if the instance is with closed registration and others are not supposed to create accounts there.
vilunov··on CT scans of coffee-making equipment
Seems to me that a bunch of websites block IP addresses with tor relays (not exits), and they all send a `x-cluster-name: eu-west-1-prod-hosting-red` header.
vilunov··on CT scans of coffee-making equipment
Does anyone else get 406 Not Acceptable openresty error on this website?
vilunov··on Making life (even) harder for proprietary modules
Linux being a monolithic kernel and drivers being a derivative work is an important design choice, which arguably is an important reason Linux became so popular now, as it forces everyone to contribute to the common pool, if they wish to use others' work. One group deciding that they are an exception is morally wrong in my eyes, because it undermines that idea.
vilunov··on Unix is both a technology and an idea
It's a system inside a system shipped as an app, isolated from other apps and with no integrations with the wider operating system. It's not far from WSL or a VM.
vilunov··on Node.js 20.6.0 will include built-in support for .env files
I personally prefer to stick to config files in whatever format is native to ecosystem (toml, yaml, ini). Env files fill the same role for apps which are configured by env variables, but you don't really need to use env variables there. It was forced by Docker some time ago as a single supported configuration format, but nowadays k8s happily accepts configs and mounts them as a file, so why jump the hoops with .env -> vars -> app?
vilunov··on IRC is the only viable chat protocol (2022)
Absolutely. Which may be a lesser problem than a closed protocol.
vilunov··on IRC is the only viable chat protocol (2022)
This may sound true, until you realize that matrix is just as extensible, and it has the same problem with clients which don’t support the full set of features.

This is not the problem. I’d say that XMPP is just as alive as Matrix, except it’s not so hip anymore. It definitely feels faster with more mature clients.

vilunov··on Ask HN: I think most SaaS can work well SQLite. Prove me wrong
Using managed technologies such as PostgreSQL will be simpler and often cheaper than SQLite, if you'll have to manage it yourself, incl. scaling, backups, updates.
vilunov··on Octox: Unix-like OS in Rust inspired by xv6-riscv
Often the user-space API limits the design of internals and available features. I for one would like to see the classical hierarchical filesystem gone, but POSIX (and UNIX) at its core is about the filesystem.

If you need to run the existing programs, feel free to develop a translation layer, at the same time allowing a new type of OS paradigm to emerge.

vilunov··on The Messaging Layer Security (MLS) Protocol
Protocols have many aspects, and the serialization format is often the least important among them, although it definitely attracts most attention, as it's the easiest to get acknowledged with deeply enough to be able discuss it.

XML back then (late 90s-early 00s) was quite an ubiquitous format, more popular than JSON. Websocket itself was drafted in 2010, so you wouldn't be able to come up with something better than a "never-ending stream of XML/JSON messages".

The single objectively positive aspect of XML is namespaces. It is very important for "extensibility" part of the protocol, as it allows you to introduce custom elements, which are namespaced by your org's domain name. Protocols such as Matrix [1] and ActivityPub [2] use JSON as the serialization format, but they too support extensibility, and therefore they have to implement namespacing. Which they do, but very differently.

[1]: https://spec.matrix.org/latest/#namespacing

[2]: https://www.w3.org/TR/json-ld/#the-context

vilunov··on Meta’s Microservice Architecture [pdf]
Placing a network call in your system also introduces:

- distributed systems problems - think consistency and ACID

- increased refactoring complexity - you now have to change the api three times instead of 1 to deploy the change with 0 downtime

- requirements for more generic CI/CD pipelines - you will develop many microservices, it's important to do it fast

it's not "just" a network rpc, even technically

vilunov··on Linux Namespaces Are a Poor Man's Plan 9 Namespaces
Or it could get translated to the target ISA by the user-agent Rosetta-style, suffering the same warmup issues, but only on some architectures.
vilunov··on The SSL certificate issuer field is a lie
It already [almost] happened twice.

Kazakhstan had plans to introduce country-wide MITM with certificates signed by a government agency, whose root certificate would have had to be marked as trusted if you wanted to use internet [1].

Many Russian government websites including the largest retail bank had their certificates revoked due to sanctions, which forced the government to start their own untrustworthy certificate center, whose root CA is trusted out of the box by Yandex Browser and has to be trusted if you wanted to access these sites [2].

[1]: https://en.wikipedia.org/wiki/Kazakhstan_man-in-the-middle_a...

[2]: (In Russian) https://www.gosuslugi.ru/tls

vilunov··on Why is Kubernetes adoption so hard?
> Nix is the way to go and keeps gaining traction

Hard to take this seriously since nix-lang is the largest (but not the only) roadblock to nix adoption in my observations. Besides, it hardly provides any benefits other than an even more exotic syntax.

vilunov··on The Gemini protocol seen by this HTTP client person
Can you provide a link to that ban threat story?
vilunov··on Zig is hard but worth it
Rust hasn't grown in scope very much since its release, but in places where it has grown, particularly async and unpin, I find that there features interact very badly with lifetimes and borrowing. I am often forced to ditch borrowing across async method calls and to put everything in Arcs, even when the lifetime is well-defined and it could be easily used if it had been scoped threads instead of futures. I fear that newer features will be incompatible with older ones in a similar way, but fortunately there are no such features on the horizon.

Considering all that, I still see Rust as the most sane choice for writing native programs. I don't really want a "better C", I want to write memory-safe software with confidence, and that means static analysis or a thick safe runtime, whatever is more suitable for the use case.

vilunov··on Slide to Unlock
iOS has been "swipe up to unlock" for some time now, as well as Android vendors had "slide to unlock" lockscreens earlier.
vilunov··on Rust: The wrong people are resigning
Then some ideas of the post went over your head. Whenever there are people there will be drama, and you wrong when you lump together ordinary contributors and drama queen. A community is not a hivemind.
vilunov··on Convos: A Persistent Self-Hosted Web Client for IRC
I tried giving it a run locally and found some annoying stuff.

On first startup and login, I found myself already connected to the libera network. Of course it failed to connect, as it chose my username as the IRC username and it was taken there. Nevertheless, I don't think automatically connecting to an arbitrary server with an arbitrary username without asking user anything is the correct behavior.

My next step was to delete the libera connection, as I wanted to connect to a different IRC server. The delete button does nothing, but in the HTTP requests log you can see that you actually cannot delete the "default" connection, and there must always be one.

After playing around, I finally attempted to connect to my server, only to find that I somehow created a second connection with an id of "" (it says so near the deletion interface).

Finally I accidentally pressed "Log out" and was met with a 404 in my browser, and a "Use of uninitialized value $str in substitution (s///) at /app/local/lib/perl5/Mojo/Util.pm line 408." in my logs.

I don't mean to demotivate the devs, as I would certainly use a kind of browser-based IRC client due to lack of native IRC clients on macOS, but as it happens, it's very lacking at even most basic user stories, while at the same time supporting file uploads and video chats for some reason. Something like HexChat or Konversation or Quassel for bouncer-users is still miles better.

vilunov··on 20 years of Nix
It's not orthogonal, Nix goes far out of its way to fix compositionality issues with hacks and dubious conventions, when instead namespaces (bind-mounts) could be used to deliver expected environments.
vilunov··on 20 years of Nix
> Just because some package didn't require another package at build time doesn't mean we don't ever want to use them together at runtime. Yet Docker conflates the two, introducing all sorts of unnecessary friction all over the place.

But Docker in fact doesn't have compiletime dependencies, it needs you to specify runtime deps only. If you want to build something in Docker, you use two-stage builds and the runtime deps of the first stage become your compiletime deps.

> This is why something as simple as getting more than one process to work with each other on Docker is such an overcomplicated mess

I don't get this, why is it considered an overcomplicated mess? If you want to run several processes in one container, you just launch it with a lightweight process manager, if you want to run it in separate containers -- well, that's even easier, just launch separate containers and configure the communication between them with a network.

> Nix gets this right with better design and composable packages

Nix actually implements it worse than Docker in some sense. Particularly, the exact problem that you described:

> Just because some package didn't require another package at build time doesn't mean we don't ever want to use them together at runtime

is not solved in Nix, runtime deps must be a subset of compiletime deps.

> that's neither composable nor customizable

Compositionality is a completely different issue on which I do have problems with Docker. DAG-oriented environment building is strictly better than inheritance-oriented, but that's all orthogonal with compiletime-runtime separation.

← PreviousPage 4 of 5Next →