Jace from ZetaNews and BeUnited has written a review of the recent WalterCon about Haiku.
WalterCon: An Experience Remembered
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Ok, I’ll ask it, who’s that girl?
http://img.zetanews.com/news/waltercon/img_4.jpg
A lot more informative than what was available on tbj. So if I understand correctly, Haiku will start an Alpha in at minimum 6 month ?
—
http://6URL.com/00P
I’m afraid she’s an entomologist…
Haiku seems to be a lot more complete than I thought it was, they need to update their status site!
Well-written and entertaining. You almost feel like you’re there when reading it
BTW, why didn’t Axel turn up? As I understand it, he’s the most active Haiku developer…
Because he lives in Germany which makes getting to Waltercon a bit logistically difficult. He’s usually at Begeistert though.
@ Cartman – hehehehehehehehe
On topic:
Indeed Haiku is way more complete than their status page shows! Cool! I must say I had some serious doubts about this project when it just started out (and I think I’m not alone on this), but that has changed completely. The site and the logo looks professional, everything seems to be really thought out, some very talented programmers working on it… I’m looking forward to the first realease!
Looking at the Status page, it seems that the two most important kits (kernel and app server) are barely out of planning and are still in very early stages of development. How true is this? From the odd quick glance at the mailing list, it appears that Axel is the only guy doing major work on the kernel (NewOS development aside).
So, how long do you reckon it’ll be before we have a usable alpha or beta, ie a standalone OS that people can just-about use (like Syllable or MenuetOS etc.)? I’d _love_ to see this project succeed, but from the stuff on the site and mailing lists, it looks like at least two or three years before it’s a real OSS BeOS replacement.
It’s been pretty well noted in the lists that the status page is badly out of date. Most effort has gone into the new website, and actual development. Hopefully, it will be updated. Although the progress level is often hard to quantify.
I’m so happy every time I read news about Haiku. It wakes my hope to start enjoying my computer again rather than just using.
Thank you thank you thank you Jace for bringing an emotional article (which most are related to Haiku but not to other tech stuff). I loved reading it and can’t wait to boot Haiku off my HD =)
Love you all
For the record, I just installed a couple BeOS distros on a couple machines for the first time this week. Developer edition on thinkpad 166mhz 80 meg, and BeOS on a p4 3.2 ghz, 1 gig. It runs surprising well on a p166, except for no 2d acceleration for the neomagic card. The driver is still in active development though. I like the GUI. It’s a pretty cool little OS, but obviously getting long in the tooth.
The OpenBeos people obviously have the right to choose any kernel they want, but I believe they made a mistake by not using linux and by using NewOS. I think BlueEyedOS has the right idea. The biggest reason is drivers. You’re always going to have to tailor your machine around OpenBeOS and the odds of getting 3d acceleration are slim to none for modern cards. ATI is back to not releasing their specs again and Nvidia, is well, as Nvidia has always been.
Michael Phipps says that Linux will always be a server kernel. The thing is that Linux doesn’t have to be a server kernel. It runs the desktop decently right now and they would be free to change the kernel and XServer in anyway they deem fit to make the desktop the best it can be. Heck, change the system calls and make it linux incompatible if they way. Rip out whole sections of the kernel and modify others to make it a “desktop” OS. I wish the best of luck to OpenBeOS, but they have a long road ahead of them.
My $0.02
DarkWyrm said the appserver will not be ready this year(6 months), but might be a year off.
The OpenBeos people obviously have the right to choose any kernel they want, but I believe they made a mistake by not using linux and by using NewOS. I think BlueEyedOS has the right idea.
I believe the issue here is the micro-kernel arch. of BeOS – which indirectly influences many properties of the system and makes BeOS non-linux like. My guess would be, that B.E.O.S. can only be source code compatible with BeOS, which would be kind of insanely cool, but still not the real thing.
For the nth time, BeOS is not a micro-kernel.
I believe the issue here is the micro-kernel arch. of BeOS – which indirectly influences many properties of the system and makes BeOS non-linux like. My guess would be, that B.E.O.S. can only be source code compatible with BeOS, which would be kind of insanely cool, but still not the real thing.
I’m curious why you think that using NewOS gives you R5 binary compatibilty and using Linux doesn’t?
Guys, the initial reason for using the NewOS kernel is probably because it was developed by an ex-Be employee (at least I think that is the case) who had ideas on how the original BeOS kernel could be worked and improved.
In otherwords it probably most closely matches the real-deal.
Just a guess on my part, but if you want to recreate an OS, why not start with pieces that are already most like it?
Anyway, I don’t necessarily agree with the “linux=server” comments, but BeOS does most certainly have a different feel about it!
Example, yesterday I ripped a CD (the same one) on my BeOS machine (dual 1ghz pIII) and on my Fedora Core 2 machine (which I am using right now – dual Athlon MP 2400+). THey not only ripped faster on BeOS, but while I was ripping, I could do all my usual tasks without even noticing that the ripping was happening. When I did it with Fedora, I kept getting mysterious “hangs” in the GUI which were very annoying. I might as well have been using Mac OS 8 or something.
ANYWAY, my point is there is some validity in wanting to use chunks of code that are closer to the original BeOS design than not. I hope I wasn’t offensive, I didn’t mean to be. :/
Go Haiku!
Mike
And btw, I thought Jace’s writeup was great.
The best info about WalterCon I’ve seen to date. Nice to put faces with names as well.
And also, re: my comment about ripping, I know the software chosen makes a difference as well. I should have admitted that… lame on BeOS –> mp3, SoundJuicer on Fedora –> ogg.
Mike
ANYWAY, my point is there is some validity in wanting to use chunks of code that are closer to the original BeOS design than not. I hope I wasn’t offensive, I didn’t mean to be. :/
You’re not being offensive. My point was that there’s nothing stopping the Haiku developers from ripping the linux kernel and XFree86 apart for their own purposes. I just find it hard to believe that starting with a 2.6.x kernel as a base they can’t implement a very low latency operating system. X isn’t slow, it’s just all the cruft that’s on top of it.
To me, it just seems like they’re using NewOS just for the mere fact that it was developed by an ex-Be engineer. I hope that’s not the case and there are clear technical reasons why they are using NewOS. I’d like to hear them
I think the Haiku people know about the long road ahead of them but chose it because it was the best way for them to realize their goals.
If you first confronted an Amiga, Atari ST or even an Apple IIGS long loNG LONG ago (1980s) then you’ll know the Nirvana that is being sought. What is trying to be accomplished is almost mystical in nature. It transcends the boundaries of technology and edges into psychological issues that are rarely discussed or even understood.
Loving a machine is possible and many of us want to love our machines passionately and to be with them producing works that are not just useful but beautiful in the deepest sense – however, without an operating system that matches our “sense of life” it just isn’t possible.
Nowadays, women are far more interesting than computers – but it wasn’t always that way
So, anyway, expediency is not the answer – it does not matter if it takes 5 years or 10 years – it will be worth it, if they can “pull it off.” Think to 1994 – ten years ago – time goes by fast doesn’t it? Don’t be concerned about speed – the goal is worth every effort as well as the time it wil take to achieve it. Drivers now are not important as Haiku will change the way people think and the drivers will come.
If Haiku succeeds, the world will be a noticeably better place.
First of all, that post was a bit creepy…
But, back to the point, can you give me one technical reason why they should use NewOS instead of Linux as the kernel.
Things like…
It transcends the boundaries of technology and edges into psychological issues that are rarely discussed or even understood.
Loving a machine is possible and many of us want to love our machines passionately and to be with them producing works that are not just useful but beautiful in the deepest sense – however, without an operating system that matches our “sense of life” it just isn’t possible.
…just don’t cut it.
I’m curious why you think that using NewOS gives you R5 binary compatibilty and using Linux doesn’t?
The thing about NewOS is that it most closely resembled the BeOS kernel.
In case you haven’t noticed. There are licensing issues with the Linux kernel which doesn’t work in the Haiku development model. MIT is the preferred license with BSDL parts if possible and GPL is when it’s necessary but not preferred.
It would seem that the kernel of choice was not easy… you say Linux, but there are several BSD kernels out there as well which are better choices indeed.
Haiku don’t wanna do what everyone does and be *nix like system. I believe many have failed to understand that everything isn’t *nix or Windows… there are other alternatives and that is what Haiku wanna become.
All NIx designs are dog old, Linux is made from concepts based of science in the 80-s, Haiku is made out of concepts in 2000…. what’s tricky to understand about that?
Almost every thread about alternative OS’s is filled with comments like, “Why didn’t they just use Linux!!?”, or “Why are they making this OS when they could work on Linux!?”, or “This OS is too small, too fast, too easy to use, and too visually attractive! Why isn’t it more like Linux!!?”.
I completely agree with the conference speaker regarding Linux’s server/desktop status.
At the anonymous from sweden –
The linux kernel being GPL is irrelevant. Modules that are linked against the kernel can be binary and under any license. That includes userspace as well as kernelspace.
Haiku don’t wanna do what everyone does and be *nix like system. I believe many have failed to understand that everything isn’t *nix or Windows… there are other alternatives and that is what Haiku wanna become.
You just don’t get it. Linux is a kernel, in reality it has nothing to do with *nix. The posix system calls are what makes it Unixy. Is Mac OSX unixy? Somewhat but it’s irrelevant to the user. The point is why re-invent the wheel when all that code is already out there. Hell, use BSD but forget about those closed source 3d drivers.
@anonymous from austin
You are confused too. Linux is a kernel, not an operating sytem. Redhat is an operating system. Debian is a operating system. With your logic you should call Mac OSX BSD because it uses a lot of BSD. Listen, they can rip out or use any part of the linux _kernel_ they want.
I completely agree with the conference speaker regarding Linux’s server/desktop status.
Are you just going to throw out random comments without substantiation?
Both of you still didn’t give any _technical_ arguments why NewOS is a better _kernel_
Straight from the horse’s mouth, found on Haiku website, FAQs section:
The goals of Haiku are to re-create the experience of using BeOS as closely as possible. Linux is a large system that has become fairly well established at this point. Its kernel design doesn’t fit well with the BeOS kernel/servers/kits design. The general consensus was that using Linux as the kernel would inevitably result in extending Linux, not in recreating BeOS.
NewOS, by contrast, is a new microkernel written by Travis Geiselbrecht (a former Be engineer). Its design is fairly similar to the (modular monolithic) design of the BeOS kernel so it won’t require great effort to adapt it. Travis is supportive of the Haiku project and is willing to cooperate with Haiku developers who are extending his work.
NewOS is not microkernel either.
Wasn’t there a newsletter article about why they choose the NewOS kernel? I think so. You’ll probably find it if you dig a little.
uhhm, you’re confusing me with PeeWee. I never said NewOS is a microkernel.
I’m not some linux fanboy that is saying use the linux _kernel_ for the sake of it just being linux. I’m all for using a kernel that is appropriate for the design goals of Haiku, but I’m just looking for some technical reasons for using a relatively new, unproven kernel rather than a BSD flavor or Linux or any other kernel for that matter.
I keep on hearing generalities though, like “linux is really a server os” or “linux doesn’t fit in with the design goals”. Maybe those reasons are valid, but in my googling efforts and with these posts here I’ve yet to find any meat on the bone.
I’ll keep on googling.
Neither BeOS or NewOS are strictly microkernels but they’re not macrokernels either, they fall between the two. I think the most accurate description would be to describe them as “microkernel-like”.
In any case, is there a strict definition of microkernel to measure against?
—
Loving a machine is possible and many of us want to love our machines passionately and to be with them producing works that are not just useful but beautiful in the deepest sense – however, without an operating system that matches our “sense of life” it just isn’t possible.
What he’s trying to say (I think) is it’s easy to use and has fast responsiveness.
BeOS was designed for this, Linux (as a system) was not. It’s now moving in that direction but it’s having a huge pile of stuff thrown on top in the process ending up with a very large system with multiple layers, these will not do your responsiveness any good.
—
The design of Linux is quite different as it has a lot more in the kernel.
Also Linux uses syncronous API calls, BeOS uses asyncronous message passing, Linux will introduce a lot of complexity and a major performance issue here.
Using NewOS will have saved the Haiku team a lot of work and the end result will be both similar to BeOS in structure, compatibility and overall experience. I think using a different kernel you’d have a lot harder time getting any of those.
B.E.OS are doing a Linux based BeOS clone so it’ll be interesting to see how well it does when it’s ready.
What would you do:
Take a small, well writen Kernel with less functionality and extend it
or take some well established monolith, wich has every feature you can think of, is a mess, and cut it down to what you need.
Don’t forgt that when OpenBeOS started they had no experienced kernel-hackers.
I fully support their decision for the NewOS kernel.
Ok, thanks for some technical reasons(finally) for why openbeos is choosing NewOS. I think the 2.6.x has gone a long way in making linux a fairly good interactive system, but it’ll be interesting to compare B.E.OS and openbeos when they mature.
<What would you do:[/i]
I would probably choose the linux kernel and adapt it to suit my needs. As I said in my previous post, the 2.6.x series of kernels has gone a long way in improving interactive responsiveness.
Why do you call linux “a mess”. Maybe some drivers are a mess, but from the limited exposure to linux kernel code that I have, I’ve found it to be fairly clean and with kerneljanitors it seems to be getting cleaner all the time.
My biggest concern is drivers. Most linux drivers are GPL and so BeOS won’t be able to use them (I think). They can use them as a reference point but will have to rewrite them. And forget 3d acceleration in openbeos anytime soon. Unless they can get NDAs from those two, then there will be no 3d acceleration. And these days, ATI and Nvidia are about the only high-performance players.
There’ll probably will be 3D, just not leading edge, you can usually get docs for lower end cards from ATI at least.
By the time any BeOS distro is ready for primetime and ATI and Nvidia still have closed specs, any 3d you will be able to get for BeOS will be for absolutely ancient cards.
sure you can use GPLed drivers in BeOS/OpenBeOS.
But as they don’t have the same driver-API you have to rewrite them.
re mess:
The last time I looked into the kernel-source was with suse 6.2 so I don’t know how the newer releases are. but I’m sure it’s gotten better.
But I hope you understand my point that for first-time kernel-hackers a small kernel is a better choice than a monster (in size and complexity) like linux.
But I hope you understand my point that for first-time kernel-hackers a small kernel is a better choice than a monster (in size and complexity) like linux.
fair enough. I’ll buy that reasoning.
Although you answer satisfies the constraint of being technical, rather than just more touchy-feely stuff it’s also wrong, and therefore not much use.
The Linux kernel and Be kernel use the same type of synchronous system call API, and the same event/ signal mechanism. Asychronous messaging is used extensively in the GUI, whether that’s Be’s Interface Kit or X11 on Linux, but it’s not the essential kernel API.
The Haiku team are already likely to have a very hard time reproducing the performance characteristics of BeOS, and they’ve yet to do anything towards that, it seems that on the whole they’re taking the Cargo Cult approach of immitating as much as they can see from the outside and hoping that the rest will happen magically.
The B.E.OS / Cosmoe / etc. people have plenty of problems of their own, but AFAIU they do have a proof-of-concept build where the GUI comes up whereas today the most Haiku can manage is a command prompt.
The most serious challenge for the Linux-based Be clones is a clean implementation of the filesystem query mechanism. That’s a markedly smaller problem than e.g. implementing a decent VM subsystem, something Be themselves never managed, but which the NewOS-based Haiku must conquer to ship a useful operating system in the 21st century.
Why is there such a big concern about 3D acceleration anyway? If I want to game “on” my BeOS box, I attach a PS-2 to my TV tuner – problem solved.
The major game makers are unlikely to focus on yet another platform any time soon – many of them got seriously stung by experimenting with OS/2 in the 1990’s, and could get bitten by Linux ports now.