Submitted by It’s been just about a month less than six years since Haiku’s last release in November 2012 – too long. As a result of such a long gap between releases, there are a lot more changes in this release than in previous ones, and so this document is weightier than it has been in the past. The notes are mostly organized in order of importance and relevance, not chronologically, and due to the sheer number of changes, thousands of smaller improvements simply aren’t recognized here.
Please keep in mind that this is beta-quality software, which means it is feature complete but still contains known and unknown bugs. While we are mostly confident in its stability, we cannot provide assurances against data loss.
This is a massive release, especially if you haven’t been keeping up with any of the nightly releases over the years. There’s so much new stuff and improvements in here that it makes no sense trying to summarise them, so I highly suggest reading the release notes carefully, downloading the beta, and giving it a go yourself in either a virtual environment or on actual metal.
after a long long time….
if this is not the case, then is not ready for everyday. Last time I tried, people told me that Chromium steals user data and Firefox was sluggish
Read the release notes.
Ok, youtube seems to work.
I’m not sure if improvements are enough or not. What I don’t understand is why they don’t port a mature solution fully compatible with modern web apps and services
The core Haiku team has ALWAYS focused on the OS itself. Apps have always been an after thought. The main requirement has been compatibility with BeOS. Compatibility with legacy BeOS apps. For better or worse.
In a sense, this has been a glorious experiment with an unknown outcome.
So your logic, do not make something new, just copy other people’s work and claim it as yours?
Where did you get the “claim it as yours” from?
This is what everyone else is doing, why? BECAUSE A WEB BROWSER IS COMPLICATED SHIT.
But go ahead and spend the time and money to get something compatible with the modern crap web – 100 developers and 3 years should do it. Because using standard code as everyone else is cheating – right?
You are right that you did not “Claim it as yours”, sorry I went to far.
However, I have a dislike to all modern browsers that seem slow, bloated and have more features than I will ever need.
But my main problem is the more code you have, the more bugs and exploits there are to take advantage of.
In-fact, the more complex the web-page and features thus needed in the browser to view it, the less I find I need the web-page. It seems to me the useful content of a web-page is inverse to how complex it is.
So I guess I am complaining about web-pages as much as getting a browser.
PS. I also notice all the browsers still do not have the filtering I would like to be INSIDE the browser itself. The idea of add-ons for features that should be in the browser, while the browser has tons of what to me seem to be little use features.
Maybe it just me being an old man.
Edited 2018-09-30 16:39 UTC
Right, that’s rational. Or, for that matter, use links2 (as I do part of the time)
However, I would assume that porting firefox couldn’t be as hard as creating or improving a quite old piece of code. Therefore, an independent firefox porting project for haiku would only add and not compete with the native browser
what are the benefits of creating a brand new browser when there are several FOSS ones that works perfectly well in 2018’s Internet? Or, for that matter, what are the benefits of evolving a crufty* browser that is way behind firefox, for that matter?
I can see the benefits of haiku, combining kernel + FS + windows manager. At least it can be a really safe environment fully compatible with old machines, and with other advantages.
However, I cannot understand why investing more time in a 20-years-old browser can be justified, since it is not compatible with a lot of services and sites and you don’t obtain any measurable benefit in exchange of such sacrifice
* crufty, as “In the beginning it was the command line…”
You do realize that WebPositive uses a port or Webkit. One of the purposes of creating Web+ was to have a native modern browser without the development headaches of porting Chromium or Firefox. Porting an existing browser means you have to implement support for Haiku’s UI and API, which means fundamentally rewriting most of the browser.
Starting from scratch meant that Haiku removed the pitfalls of trying to port the rendering engine and UI in favour of just porting (very portable by design) render and building their own UI on top.
WebPositive might not be as functional as chrome, but it’s a darn sight better than BeZillaBrowser (a FF2.0 port)
Edited 2018-09-30 14:37 UTC
Then the same rationale would apply for any other port from linux, right? No pun intended
Pretty much. Haiku lacks any sort of X11 support, given it’s got its own custom GUI. POSIX support is also spotty
The QT port is fairly fleshed out, so any QT apps can be ported fairly easily. That’s how Haiku has support for LibreOffice. If it’s not QT based, you’re in for a headache.
I know at one point there was an effort to port WINE to Haiku. I believe that fell apart due to no X11 and lack of POSIX support. That was about 5 years ago, so things may have changed since then.
Qt, not QT (QuickTime)
Hm, and IIRC Libre Office isn’t only Qt …anyways, its port to Haiku still requires a bit of work – it renders fonts badly, scrolling isn’t fluent, still crashes quite a bit.
I think they are a bunch of people doing this amazing work because they want to do it, so there are not enough resources to maintain a lot of ports and subsystems.
So, if you want a port, the best way to help them is reporting problems properly in their bug tracking systems instead of bashing their work or helping to develop such ports.
There is WebPositive, a WebKit based browser.
Yes. QupZilla/Falkon is available and is a modern Qt 5 browser with HTML5 support and all sorts of modern features you’d expect. In Falkon GIT (upstream) work is ongoing on implementing a better extension system so that it’ll be easier to create extensions for it.
Edited 2018-09-29 14:44 UTC
That’s one of the biggest improvements from Alpha 4.1. Back then Web+ was a janky, crashy piece of shit totally unsuited for regular use. Nowadays my biggest complaint is that it lacks an adblocker.
Web+ still as some issues on newer websites, but this seems to stem from similar issues with.. a bit one is the cryptoqueue locking up. Which locks up the whole browser and requres that you kill it and restart it…
But other than that it is pretty nice. And thanks to updates being possible to push out via the software updater we’ll get fixes for that before the next release.
Note haiku effectively has a rolling release system going now with a stable and master branch… with stable fixes getting cherry picked back into the stable branch out of master.
Fixing cryptoqueue and even adding hosts-based adblocking would be enough for me.
I’ve noticed it’s a bit crashy on Youtube when i was playing with it in a VM
This is a big milestone! Lots of hard work went into this.
It certainly is! I haven’t tried Haiku for a couple of year now but I have a old Pentium III system with BeOS installed that I may install it on this weekend and give it a whirl.
Good times ahead!
I’m definitely going to try this out.
OK, that was fun. It crashed during boot on my laptop (Intel i5 4400U), but started up fine on my desktop (i4770, GeForce 1080, some Intel WLAN).
Networking is a bit buggy, in that it found my 5GHz network but wouldn’t connect, and then refused to discover any other networks before it was disabled and re-enabled. Then the old 2.4GHz worked. WebPositive is a bit shit, but did play Youtube with audio. At least the ad at the start — it didn’t want to skip to the actual video.
The UI is delightfully lag-free. It’s as if it anticipates my clicks, and opens new windows exactly as I hit the button the second time on double-clicks. The responsiveness of BeOS is all there, and even better now with multi-core CPUs of several GHz and SSDs. This is of course where computing history decided to take the wrong turn, and go for translucency and windows that fade in and out instead. It would be interesting to see Haiku gaining some traction.
If you want a better web/YouTube experience, install QupZilla/Falkon from the package manager.
My modern testbed for Haiku is an ASUS 1015e netbook that originally shipped with Ubuntu. It has UEFI but no Secure Boot, and supports legacy BIOS/MBR booting, so it’s a great machine for testing non-Windows OSes.
The Haiku beta 32 bit hybrid ISO had kernel panics every time I tried to boot it. The 64 bit ISO booted fine both in EFI and BIOS modes, and it installed fine in both. It wouldn’t boot from the hard drive in EFI mode even with a correctly set up ESP, but BIOS mode booted normally after install (I had to format the drive for MBR of course).
This machine is technically underpowered by modern standards, with a 1.1GHz Sandy Bridge Celeron CPU and 2GB soldered RAM, but it’s actually a bit of a sleeper. Hooked up to a larger screen, keyboard, and mouse, it is more than fast enough for everyday work, as long as one doesn’t expect to play 3D games on it. With Haiku, it’s insanely peppy and feels like a new computer. It blows away even the lightest Linux distro in terms of responsiveness, file access times, and startup/shutdown.
I had some rendering issues in Web+ on certain sites, and for some reason my 5GHz Wi-Fi connection wouldn’t automatically reconnect after a reboot, but so far everything else works well. I’m looking forward to 2D accelerated Intel graphics drivers; it seems that and the webcam are the only hardware not 100% supported, and I don’t care about the webcam apart from completeness.
Edited 2018-09-30 13:46 UTC
You could install Haiku on a Pentium II and it’d feel peppy.
I ran BeOS R5 PE on one about a decade back. Made my Win7 machine look like a sloth
Booting Haiku on notebooks can be problematic due to the integration of components and firmware.
Entering “safe mode” during boot by holding the Shift key pressed allows one to start figuring out the reason(s) for failure to boot.
If the boot loader reaches the Haiku splash screen, there is a high probability that entering “safe mode” will help. However, it the system hands before the splash screen is displayed, then it is a lot more complicated.
I’ve had Haiku R1A4 booting on a macbook, with Ethernet (no wifi), graphics and sound working OOTB. I don’t think there’s anything more integrated than a mac.
OTOH there are not that many Macbook configurations, this helps with support.
I’d like to see some effort put into supporting macs OOTB, with full media-key support, hardware support etc. Their hardware configuration is fairly generic, and it uses integrated intel graphics, which haiku already supports well.
And no, don’t bother adding support for the touchbar.
Hangs on boot
How do you get coredumps or whatever it’s called?
Depends on where it’t hangs.
First place to look is boot up, and hold down shift while booting … this will get you to the debug menu where you can try and figure out what is causing the issue.
I suggest you boot up hit shift, then from there enable the on screen debugging and also disable paging.
then continue booting you’ll get debug output on the screen and that might help…. if the screen goes blank at the end or gets garbled you probably have a video driver issue, in which case you back blacklist the accelerant for your card by browsing to it from the lacklist menu and selecting it. Failsafe video is also something you can try that’s a litte easier but in my experience doesn’t always work.
Also if your monitor going out of sync during boot make sure you select a resolution it can display before continuing with the boot!
HAve fun and if you find a bug in Haiku make sure you report it or at least check that it is already reported! Thanks
The icons that run under the logo indicate what stage it’s loaded/crashed at. What’s the last coloured icon?
I’ve just tried it on Virtual Box.
Reminds me of the distant times. When BeOS was my main OS.
What seems to be missing now is software. There is not much in the packet manager, and old sites as BeBits have vanished.
I’ve tried the software I wrote 17 years ago, it hangs when I press the “Config” button, wonder if the bug is in Haiku or in my code.
Old software is probably trying to put stuff in config folders that are no longer writable due to the package management integration.
Honestly this is a small nit I have with how the did it but so be it. I personally think they should have made packaged subfolders for the packaged read only stuff instead of making non-packaged read write folders whee you can place data that gets mapped into the normal hierarchy.
This definitely is a concern. However, if the application didn’t hard-code paths (and used the proper API calls to find the config paths) it shouldn’t be an issue.
If the normal paths were read write it never would have mattered… but water under the bridge most stuff works now anway.
I brought up this issue back when it was still in dev but apparently there were reasons not to do it that way.
There are two Haiku variants and it is uncertain which one you used.
The 32 bit one is designed to be compatible with the old BeOS applications – albeit one may have to adapt them for installation due to the introduction of the package manager and sticker rules for where an application is allowed to put its configuration and helper files.
The 64 bit one is not compatible with the old BeOS applications at this time. This may change when the 32 bit compatibility layer subsystem is completed.
As for software, while the old BeBits and related Hakuware sites have long disappeared, there are other sites which have taken over – for example BeBytes.
The list of available software through package manager is quite long. Not all of it shows up initially.
Both Puri and Documentviewer are available in HaikuDepot, the new package manager.
I guess I’ll need arctic clothing for when I’ll go after death to hell.
I used to spend a lot of time with haiku (and I hope the programs written by me, Puri and Documentviewer are still working), and I appreciate their work, but if I had to make an educated guess, I would concentrate all the effort on risc-v and vulkan support.
When you are small, with little man power, starting early into this direction is a risk, but it may pay out.
X86 to me doesn’t look like the future, for sure not for me. ARM attacking risc-v clearly shows that they have a reason to be afraid.
cipri
x86 is the architecture of the computers people own now.
RISC-V is nowhere, except in rare eval. boards or FPGAs. Mostly embedded stuff. Haiku is an OS with GUI environnent.
If Haiku can be compatible with a few cheap x86 laptops, (some are really discounted because they are barely usable with 32GB EMMC on Windows), it can have far more audience. And there is virtualisation.
And, despite its many kinks, x86 is a relatively open architecture, supported by many open source software.
There is no guarantee that RISC-V will get out of the embedded space to reach desktop and laptop computers. ARM, despite its huge presence, have not managed it, except for a few Chromebooks and Windows RT netbooks.
Edited 2018-09-30 20:58 UTC
entering a market or domain and to compete with somebody that has already a lot of experience and assets in that domain is a lot harder, than both starting from zero.
In the electric-car domain, tesla could compete and win over mercedes. If they tried entering their old business, I guess they would have be by now just a small company. They would have needed many years to catch up a little. Now if you look at tesla, you have to think if mercedes will manage to catch up with tesla the next 5-10 years. If I had to bet, I would bet on no. I think in 10 years, tesla will sell more electric cars a year than mercedes. If you check how many big companies are behind risc-v (google, nvidia,…), it is hard to think, that it will fail. It is already a success, just it is not yet very visible. The Freedom U540 SoC, already today looks like acceptable to me, and I am pretty sure, next year, I will have a risc-v machine, running at least a major linux distribution on it.
Haiku has of course a lack of kernel developers, that’s why a port of haiku to risc-v is perhaps something we can not expect at all. I prefer to work on something that give me the impression to be meaningful in 5-10 years, than working on something that has good chances in 10 years to start loosing market share, and disappearing slowly.
Oh man! I wondered how long it would be before cars and a comparison with cars came along.
Haiku isn’t a car, electric or otherwise!
Yes, Haiku is not a car. And tesla is not a cpu.
Still it was a general idea, just with an example to better understand the idea.
And there is a connection. Tesla lost recently top chip architects to risc-v startup esperanto technologies!
https://electrek.co/2018/08/20/tesla-loses-chip-architects-autopilot…
As in most open source projects, patches are welcome, and I’m sure a RISC-V port would have no problems being accepted as such
Tesla and Mercedes hardly compete with each other – the latter makes mostly trucks, buses, industrial engines; passenger cars are a side business for Mercedes.
(that said, I agree with your general sentiment, as I expressed in recent discussion about Sculpt/Genode http://www.osnews.com/comments/30749 )
PS. Was there ever more recommended OSAlert story than this one?
Edited 2018-10-02 13:17 UTC
Totally agree Haiku needs to target a niche if they are to have a hope of being mainstream eventually. However all niches are determined back from the customer, not forward from the tech.
Personally I like the idea of a simple out of the box privacy first machine built form the kernel up. I can do this for myself, but only with a lot of research and not a lot of confidence. It’s becoming a huge need that none of the big players want to happen. So someone new has to do it. Maybe collaborate with DuckDuckGo to get some browser coding horsepower.
Privacy starts with safety, software safety, software safty can build up just on hardware safety. If you don’t have guarantees of correctness, software alone has a hard time to guarantee safety or privacy. That’s why it is crucial to have an open isa, and to have open hardware, and risc-v is by far the best option we have today. Correctness is very hard to get, and the seL4 micro kernel tries to be mathematically proven as correct (implementing the specifications bug free).
You heard about Tesla, but did you hear of Alan Cocconi and AC Propulsion? I guess not. There it started the journey of tesla, they borrowed their technology and decided to make on top of it a high performance sports car. Ac Propulsion themselves decided to make a little ugly electric mini car on top of it (https://en.wikipedia.org/wiki/AC_Propulsion_eBox). Tesla decided to enter the high performance “niche”, now you see how history turned out in favor of tesla.
Targeting performance has better chances than targeting obsolete systems.
You are effectively referring to blue ocean strategy. You don’t just enter a market with a single feature. You need to create an asymmetrical strategy that is amazing at a viable range of highly desirable traits and in turn you will be worst at others. Teslas are beautiful, luxurious, self driving, up-gradable over the air, wicked fast, super safe (if you stay awake) and a status symbol, as well as electric. Also highly expensive and could only go for a 300 miles with a charge time of 1000s of times longer than gas and only at highly scarce (compared to gas stations) charging stations.
I was thinking Haiku could be refreshingly simple and super fast as well as private as well as self optimizing given the highly aggregate thread structure. However the goal would have to be address the needs of the niche, not revive ancient OS tech to the letter or run on everything. So you are right. Also agree the code needs to probably be open as you cant trust corporate these days. BTW Tesla effectively revived ancient car tech. Electric predated the internal combustion engine.
I’d like to see the ARM port get more traction. I’d love to be able to run Haiku on a Raspberry Pi.
It sounds as though the base idea for the package manager may be similar to Nix, though simpler and not offering as many features, perhaps. Are there any articles or discussions along these lines? Are there plans to add more Nix features in (or maybe, allow Nix to be used directly)?
No it isn’t like Nix.
It’s more like each package is mounted read only into a read write filesystem and there are non-packaged folders also strewn about where you need to modify a file that is part of the package that “shines through” or overrides the packages contents.
Dependency wise those are all generated at build time and manged in HaikuPorts. Then dependencies are resolved at install time from the repos or available local packages.
I encourage people to install Haiku a flash drive and boot from that. This way you will be closer to the metal and feel the speed of very streamlined OS. I would also encourage anyone who has written an application to check out the online copies of the BeBook: https://www.haiku-os.org/legacy-docs/bebook/
Writing C++ code with this incredible API is really a programming joy that must be experienced first hand.
How lovely UIs still were before they introduced ‘flat design’! Hope they don’t change that.
And no need for ‘themes’.
Genuinely hoping this results in an awareness boost of the OS & spurs some developers who may have been watching from the wings, for sometime, to finally see it’s worth & contribute. It deserves more attention & hopefully has a brighter future now than it did a few years back.