VitruvianOS
51 points by blainsmith
51 points by blainsmith
TL;DR it's a Linux distro with some additions to allow running haiku applications unmodified, packaged with haiku applications.
https://v-os.dev/news/vitruvian-0.3.0-available/ seems to be heavily AI-edited, so make of that what you will
I am not seeing the tells that you are seeing.
What am I missing?
As a huge fan of BeOS who isn’t interested in running Haiku in a VM or acquiring ancient and specific hardware to run it natively, I find this to be a really exciting development.
This makes it a genuine shame to click here and see the sole comment to be an insubstantive accusation.
I’m no AI proponent but at this point I’m fairly certain my own writing style is at significant risk of being dismissed, so I’ll admit to taking your comment a bit too personally.
What I saw:
Now sure, any one of, or even all of the above, is not 100% indicative of AI writing. Still, taken as a whole, the piece exudes the same "voice" that is so common among AI models nowadays.
On a grammatical & informational level, the writing is fine. Great, even! But I am tired of that voice. If you (generally, not you specifically, with your human mind, choose to put words to page yourself, the end result will in all likelihood sound much different, which to me is a positive signal that you care a lot about the project.
Now to be completely fair: the About page & various samples from the wiki only have excessive em-dashes, and no other prevalent tells, so my read could be wrong. In any case, I'm sorry to have been unnecessarily accusatory.
I agree that AI is exhausting, and my eyes often glaze over when encountering That Voice, so I sympathize with your frustration.
In this case though, I just read it as a small development team trying to use the traditional convey-and-induce-enthusiasm language tools. If AI was involved, it reads like a milder case to me (like when a non-English speaker uses it to polish their language, or likely in this case a non-marketing person uses it to polish their copy).
Regardless, thank you for your explanation and for engaging in such a reasonable way. AI sometimes feels like an intentional DDoS on our collective patience.
In this case though, I just read it as a small development team trying to use the traditional convey-and-induce-enthusiasm language tools.
It calls into question the nature of the work done and how much was done by AI. That compromises both the licensing of the work and also the quality.
Mere suspicion based on the use of a few inconclusive indicators in marketing copy should not be allowed or promoted to cast such a shadow over an entire project.
The implications you are describing are precisely why I took issue with the original comment. It should be considered uncivil at the very least to cast shade like this.
Personally my attention is drawn to the high volume of fluff:
The result is a desktop that doesn’t impose itself between you and your work, but instead enables it.
Unlike traditional operating systems that prioritize features, services, and monetization, Vitruvian asks a different question: what would I actually want to do with my computer that I currently can’t?
It’s that question that drives every design decision.
Vitruvian is not an operating system like others.
There is very little information in this excerpt, and this is the style that most LLMs output by default. They can definitely be coaxed to output different styles with appropriate prompting and examples, ChatGPT and downstream/distilled models all seem to generate stuff like this.
edit: that said, maybe I just don't like this writing style and I'm holding it against the author. don't let some random person online tell you what to do with your project /shrug
or acquiring ancient and specific hardware to run it natively
I have Haiku installed bare-metal on a Ryzen machine, and have since I got it years ago. Other users have it likewise running on recent AMD and Intel hardware. The new (experimental, still out-of-tree) NVIDIA acceleration drivers work on Turing and up. So, "ancient and specific hardware" hasn't been required to run Haiku for quite a long time.
This is hopeful news!
Are you using VESA for graphics? What is the experience of that (or whatever solution you are using) on a 4K display?
Nvidia drivers are unfortunately of no use to me (never have, never will).
EFI framebuffer. Works fine on my 4K monitor at least.
Thanks for this. I didn’t note your username previously but I most definitely recognize it from all my perusing of Haiku status updates over the years.
I will give it another go on some spare metal soon. In the meantime, I’d like to issue a huge thank you for all the work you have put into keeping the dream of Be alive.
Honestly, if what LLMs spit out were anywhere close to that page I'd think a lot higher of them. In my experience the signal to noise ratio is typically a lot worse. Of course it states visions and "values" to some degree, but I'd expect that from a project like that - in terms of age, purpose, popularity.
Also if it really was just basic editing of a text and not for "write me an article based around those cues" it's a different story. In other words, it LLMs were involved I think "heavily AI-edited" is a bit of an overstatement here. Also I get it when a non native speaker or simply someone who isn't a good writer wants to have an LLM have a pass over it. I think there's a difference between someone who essentially "cannot be bothered" and someone who might be vary about how to best write/format something. It's not like that person was paid to write the article or the article itself being the main part of the project.
So again, if people would use LLMs like that I'd think a lot better of them. The sad reality is it's usually used as a tool that allows companies to circumvent copyright and licenses for profit, usually at the same time claiming copyright on said work.
I swear, nothing kills my enthusiasm for something like this faster than that.
Was going to show someone this post because I thought they’d be interested but I respect them too much to bother them with what might be slop all the way down.
I'd love to have a more Haiku/BeOS style Windowing Toolkit take hold in the Linux/Unix world. I think the QT and GTK "duopoly" is rather bad. While there are other contenders that also seem great I think some of them end up getting closer to those and personally, subjectively I think that's bad and certainly not the best way to do widgets, both from a programmer's and a user's perspective.
And since this often comes up regarding new developments in the field: While I understand accessibility is important (I have been working and promoting it and used it as a counter-argument for Flash back when people would consider it the future), I also think that any project has to begin somewhere. So while I think it should be kept in mind and certainly not be postponed indefinitely it depending on the project might also not have the highest priority before a project even has users.
Has someone who has never used BeOS, could you explain what you miss with Gtk/Qt?
So…windowing-wise, I don't honestly miss Be that much. The only thing it did that I do kind of miss is that you could actually drag the tab around at the top of windows, so you could kind of manually stack up related windows and tab between them quickly. Think browser tabs, but for arbitrary app windows. The downside was that (unless I'm misremembering?) there wasn't really any good window organization stuff (no tiling, no snapping, etc.), and I don't even remember for sure if you had window minimization v. a Mac OS 8-style collapse-into-the-menu-bar thing.
Haiku heavily improves on all that: you can snap and tile related windows together, and they move as a unit, making it easy to group arbitrary apps together. You can definitely minimize/hide windows. You can resize on any side, and you can trivially send a window to the back or drag it forward. You can see the full list of capabilities if you're curious.
In terms of the toolkit: as a dev, IMVHO, it's fine, but that's all I'm giving it. Haiku and BeOS are much faster and lighter-weight than Gtk, but you also get a lot less. Be designed their APIs in mid-90s C++ OO style, which means they're subject to fragile base class and compiler ABI issues. The windowing toolkit is best-in-breed 90s-style event-loops, but that's it: you get a rich event object, but you're not getting signals and slots or anything like that. Layout is entirely on you, and there is no designer.
Haiku specifically improves some of the above (e.g., there's a fully layout manager), but it can only go so far without breaking BeOS source compatibility, which at least has historically been a no-go. I personally honestly like GObject and Gtk pretty decently, and definitely find using it much more pleasant than Haiku's APIs, in part because things like GIR mean I can fairly easily use them from any language (not just C++), and in part because, hey, I have a convertible tablet, having a UI that's designed to be touch-friendly genuinely is pretty nice.
As a user, though? The speed is great, everything looks like an honest-to-goodness button, the scroll bars are visible all the time and very obvious, drag-and-drop everywhere is a thing, etc. If you're in the system, it feels cohesive and coherent in a way that not even Mac OS 8 really every managed to achieve.
So, as a dev, I don't personally miss coding BeOS. As a user, I do miss the experience.
I'm thinking the compiler ABI issues with C++ are a non-problem for Haiku at the moment since approximately every single app that targets Haiku is open source and can be recompiled anyway?
They were still an issue last I looked. Haiku has a really awesome app distribution system, but it's distributing binaries. It wouldn't work well if they weren't mindful of ABI issues, because apps would constantly be breaking as they expand the system. Now, there are well-known workarounds for all that (e.g. pimpl pattern), but I don't think Haiku went as far as retrofitting that type of thing onto the API.
More with GTK then QT. Simplicity. Many things seem to be in the way, undescriptive, unintuiti or simply horribly complex. I think we took a big step down in terms of usability compared to the late 90s and 2000s.
One of the most annoying things is GTKs File Picker. Everything one could possibly want from it is hard. It's hard to see the full path, its hard to get to different places. It's hard to know whether you will pick I or open a directory. And while the only thing I do like is that there are bookmarks it can be hard to create them, hard to distinguish them (god forbid I have the same directory name in two different places, it's really rather slow, creating new files or directories within the picker is confusing at times. And lastly the navigation, which I would argue is the main task is also annoying. Good luck typing or pasting a path you already have a hand. In a world where people live in browsers why would you remove the URL bar for you local system?
I know these aren't things that I want from BeOS but I want both fresh air and something that makes sense and it's consistent.
I don't mean the huge parts. There's also the general desktop bloat part that Desktop Linux very much embraced.. At least there I get where it's coming from, I get the rationsli and while I think it's really sad that road was chosen and not more of a middle ground, I understand why it came to be and it's more of a philosophical/opinion/taste question. I understand why everything now goes over xdg utilities, etc. To me it's too bloated and I think the whole "expect everything to be like KDE and Gnome is a bit annoying, just like dependency on systemd ("it's just an ibit system" and "you don't need to use it" is probably why now every DE and large swaths of software now have a hard dependency)..
There are more ways to do things. And being a half baked OSX isn't what I want desktops to be
I know that was a bit of a catch all but GTK and QT are huge. Don't get me wrong, I think they do great work overall, a lot is personal preference and I'm certainly not one who dictates how things should be.
But all of that makes me happy when people try to break out of that. I like it when there are other approaches to problems. I dislike it when people act an approach by someone with enough funding, time, marketing or the most polished way is somehow labeled as the only/best one.
I think it's really cool where QT and GTK and Gnome and KDE interoperability have gone. I know a big part of that is possible through these things. However in a way that seemingly demented it as "the only way" so to say which in my opinion is a bit of a shame.
Just to reiterate. I think these projects are great and if I didn't think so I'd spend a lot less words criticizing them. And since I get the "why" I don't want to blame the projects or anything but see it more as a "how it came to be".
I hope that somewhat answers your question. But more directly: I want more approaches and the BeOS/Haiku way is one that feels very much underrepresented. Same with things like QNX, and many other approaches, there's projejects
The thing with these frameworks is that they offer a lot which is great. But since they also basically dominate Desktop Linux (which certainly isn't an accusation, well done!) they dictate many things and today in a degree that simply influences everything else, because if you don't provide this or that mechanism in one case it's just weird in another borderline unusable. Practical example: GTK (something like Gajim) and system trays, which are a thing that from a user's perspective always just worked. Even in obscure systems it was not hard to have a tray icon. (SNI vs xembed). That's just one example.
It's odd but interoperability (and of course new features) in a way lead to "everyone else" be a bit left out.
I honestly don't see the value in Haiku applications and desktop without the Be filesystem. You might as well use KDE. I'd love to be convinced otherwise.
What are the advantages of the Be file system? I belong to the group of users that doesn't even know by heart which filesystem they're using, let alone which they would prefer.
I thought the idea of BeOS was a complete ready to use system for your personal computing needs. Of you need to fiddle with instalation of this and than piece of software on demand, then the advantage is already gone.
How recent is your experience with Windows search functionality? :)
It’s worse now on Windows 11 than it ever was, but it’s very hard to demonstrate the visceral distinction between searching for files in BeOS vs Win98.
BeFS was crazy responsive in an era where that wasn’t the norm. But it really shone in its leveraging of attributes. Default applications were defined according to MIME type, not extension or resource forks (or whatever MacOS 9 used).
Users could define arbitrary attributes and surface them in the file manager (Tracker). MP3 ID3 data was available as columns in Tracker. Emails were stored as text (but with an email MIME type, IIRC) but sender, subject, etc were all available as columns. (This was the POP3 era, when email downloaded from the servers).
You could create “folders” which were really just queries onto the underlying file system. All emails from so-and-so, or all songs by such-and-such, just one click away and fully integrated in the OS from a UX perspective.
Oh, and these queries were active. Downloading a new such-and-such song from the BeOS Napster client would immediately surface that song in the query.
Personally I was surprised that BeFS wasn’t the file system used here but I’m not married to it by itself. I just want the experience again. It’s been 25 years and nothing has come close!
Edit: For more information, check out Practical Filesystem Design. It seems to be publicly available at this point. The designer of the FS is the author of the book.
Emails were stored as text (but with an email MIME type, IIRC) but sender, subject, etc were all available as columns. (This was the POP3 era, when email downloaded from the servers).
Just to add to that: the MIME thing wasn't just a quirk, MIME types were pervasively as the standard identifier for a bunch of things. Including applications, for that matter -- each application could declare its identity by defining its own MIME type, which was then used as an identifier for message passing. BeOS's entire scripting interface was amazing for its time.
The screenshots look fantastic! I’ve never used beos, so I can’t comment directly on the OS itself but the UI, with its elements of depth and shadow look so good now. I think the current UI design trajectory of stripping everything down to blindingly white flat surfaces is beginning to lose its luster. I see this UI and I feel like I can reach out and touch it.
Interesting project! This might solve the problem of "not enough drivers" for Haiku?
Tangentially related: I "recently" installed Haiku on an old computer from ~2010, bare metal, and was pleasantly surprised by how incredibly snappy it felt. And as I started peeking through the software repositories, I was equally surprised by the amount of packages I saw.
Haiku is such a delight to use. The modifier key window interactions are a little intense, and its system of file system attributes is probably doomed to never be portable, just like macOS lacking file extensions was a compatibility problem. But I hope this helps the Linux GUI world take away both lessons and code.