It's funny, I just wrote up some experiences with the latest Haiku beta and also ended up missing applications despite the snappiness and amazingly trim fit into much less than 1G RAM. I'm sure many more apps will come soon though.
It's worth noting that at this point in history (2001) Mac OS X is still a turkey; visually unique, but very slow and incomplete. It would have only been of interest to the freaks and geeks in the Unix/OpenStep community. I can only imagine there was much hand wringing in the old school Mac community about whether Apple was on the right track or not.
It wasn't until version 10.3 in 2003 that OS X rounded into a form that would drive the final nail into the coffin of Classic Mac OS.
In 2005, version 10.4 takes a giant step forward, leap frogs everything else on the market, and becomes a genuinely compelling alternative to both Windows users and Unix/Linux users. This is also the version that adopted arguably the only really compelling feature of BeOS (Spotlight), and the first version to support Intel.
arguably the only really compelling feature of BeOS (Spotlight)
I’d argue the most compelling feature of BeOS, one that to my knowledge has still not been replicated anywhere else, was its amazing UI fluidity. Seems like it didn’t matter how bogged down the system got, the UI was flawlessly responsive and boasted latencies today’s desktops could only dream of.
One particular thing that was kinda nice was a separate MIME type attribute for all files, fully distinct from filename (and specifically extension), and used consistently throughout the OS to determine available actions etc. Although the real magic here is the latter - providing storage for an extra attribute in the filesystem is easy enough.
I was a Linux sysadmin and was quitting my job to go to college. I had been using Linux on the desktop since 1995. I thought I'd buy a laptop, and why waste my time trying to worry about hardware compatibility with Linux when I already have a Linux desktop?
I figured, "let me just buy a Windows machine so I can easily play DVDs and use Office". When I stepped back and realized that was my only criteria, I thought, "why not try a Mac?"
I brought my white G3 600mhz iBook to (2002 tech company) work shortly before starting school, expecting people to make fun of me for buying a Mac ("everyone" used Linux, but Windows was begrudgingly tolerated in the community) and EVERYONE was interested in OS X.
I used that iBook for tons of work stuff in my last couple of months. Popping a serial console on SunFire 6800s, directly connecting via ethernet to them and forwarding X Apps to the Mac, etc. It was a revelation.
Not only that, but you could actually close the lid, put it to sleep, and the battery wouldn't be dead when you came back from lunch.
I had an iBook G3 that I used with Linux PPC. I'm not sure about your model, but I think that entire line had hardware suspension, so it had little or nothing to do with OSX.
It was one reason I got it, fast suspension was all over the place on PC laptop hardware on Linux at the time, but on Mac hardware it would Just Work, as long as your OS supported it (which I think amounted to just telling the firmware to initiate suspension).
The other reason being that battery life on PPC was far superior to x86 at the time, that was one of the things that got a bit worse with Apple's x86 transition, and better again now with their new M1 hardware.
FWIW I preferred OS X to Windows 2000 and Mac OS 9 and used it as my daily driver starting with the public beta. But maybe I wasn't a very demanding user since I mostly used the terminal, OmniWeb, and Mail.
I ran Cheetah and Puma on my 9600/233. I tried really hard to convince myself that it was useable. Though to be fair the UNIX underpinnings were a revelation and it felt like living in the (gorgeous) future. Just a very slow one.
I then used a B&W G3 until it went out of support. It indeed became better with every new release, which was quite impressive.
Every release had a ~2 factor in performance improvements between the public beta and Tiger (10.4). Performances degraded a bit with Leopard (but improved again with Snow Leopard). That’s why Tiger is a bit of a high water mark for PPC machines. It was very useable on a G3 from 1999, albeit with a more recent graphics card for Quartz Extreme and quite a bit of additional RAM over its initial specs.
10.4 brought Spotlight
10.5 introduced Time Machine
10.6 cleaned everything up and added some stuff like Exchange Support. IMO the peak of Mac OS X.
From then on, Focus was put on Social Media integration and data collecting services.
Seems I‘m getting old an nostalgic.
It's all subjective, but I'm not sure why they drew the line in the sand at 10.4. 10.2 was the first release that had hardware accelerated 2D graphics, which dramatically improved the UI responsiveness of Aqua. By 10.2's release Apple had also discontinued the development of Mac OS 9.
The OS X Finder was unusable until 10.3, and ISTR there were still other missing things from OS 9 that 10.4 finally rectified, but I don’t recall what exactly. I do recall recommending against relatives and friends installing OS X until 10.4 finally filled in the gaps.
edit: you can argue with “unusable,” but I never saw anyone use the NeXT Workspace or the OS X Finder pre-10.3. Everyone on those systems used the command line to get around.
I think you’re exaggerating. I was a student at a time, started with Mac OS 10.0 on a beige G3. It was slow but it worked. I browsed the web, did my homework, chatted with friends and used Classic for apps that we’re not converted to Carbon yet. It used the terminal because I was amazed I could do so much (like run my own web server), but it wasn’t a replacement to the Finder.
Didn't 10.4 run only on some special developer Apple x86 board? I know it was then successfully appropriated by the hackintosh community, but I don't remember an x86 Mac being released with Tiger
I ran BeOS off and on from the Intel transition to the end, I still regret not joining the party early enough to experience their custom hardware.
I'd say mostly not.
Sure, you can browse the Internet and play music at the same time on any platform today, but a lot has changed.
Much of what was good about BeOS was breaking with the past and starting from zero based on today's desktop needs, a modern take on the Amiga in many ways.
Existing platforms may be able to steal an idea or two, but that's different from being built from the ground up to solve a specific class of problems.
To begin with, absolutely everything was multi-threaded, there was just no way around it.
Fast file system queries that could be stored as shortcuts were nice, still haven't seen that implemented nearly as well elsewhere.
I wrote some software interfacing the system C++ API's and found everything very convenient, clean and tidy compared to any other platform I've programmed.
Fast file system queries that could be stored as shortcuts were nice, still haven't seen that implemented nearly as well elsewhere.
NTFS metadata architecture really inspired by BeFS, or at the least, a fellow traveler.
You can tweak the Windows Explorer (the filesystem UI) to operate on arbitrary metadata. Well, I was able to get an email inbox, where each email message is a text file, just using Explorer the BeOS way, on Windows 2000. It wasn't exactly fast.
> Question for you OS geeks: did we take the good things learned from BeOS and implement them in Windows/Linux/macOS?
Somewhat of an aside, but I wish there existed a kind of "encyclopedia" of operating system features (both UI and otherwise, with a special emphasis on "dead ends" like BeOS), that pulled together demos and documentation into one exhaustive reference.
It seems like a lot of good ideas get lost is the sea of lowest common denominator solutions, and an ecyclopedic reference would make it easier to resurrect those ideas.
I tried to start a talk series last year about these dead ends, but only https://media.ccc.de/v/rc3-525180-what_have_we_lost actually made it out. Perhaps I'll pick up the thread later, because there are a lot of dead ends that had some amazing ideas that it either wasn't the right time for (BeOS, DomainOS), bad business decisions (Genera), or particularly insular communities where they were used (OS/400, TRON)
Not much, unfortunately. File system metadata query and responsiveness is something that I haven't seen matched still. The latter is an exhilarating feeling that I'll never forget. Also, the UI had a “fun” but still elegant feel to it that I was very fond of.
The below is based on vague memories from 20+ years ago...
Some examples:
Every email received was stored as a file. All metadata about the email (sender ,subject, headers, etc) were stored as attributes.
Tracker, the file manager application, could display any arbitrary attribute. I seem to remember that instead of an email application's window, my inbox was a simple Tracker window showing the specific attributes required to know about my emails.
My music collection was stored and displayed the same way.
Everything was searchable but also usable. File attributes weren't hidden behind CLI commands and flags like xattr, they were part of the GUI and every application using the SDK would know/understand them.
This was, at least in part, how Microsoft's "Internet News and Mail" worked. Not sure where it kept metadata, but the initial implementation seemed like an Windows Explorer extension, which was a very clever approach, IMHO.
I wish Gnome did an e-mail client like that, synchronizing a local folder with a remote IMAP mail store, or presenting a remote IMAP storage as a mounted folder under an IMAPFS sort of thing. Arguably, that's easier to do with POP3. Same thing with an outbox, where writing a properly formatted file would send an e-mail.
> File attributes weren't hidden behind CLI commands and flags like xattr
This is a design choice in the UI, not something intrinsic to the OS.
BeFS maintained indices of extended attributes; in other OSes indexing is done in userspace which is safer and more flexible but can introduce race conditions between filesystem operations (e.g. moving a file) and updating the index.
Most of what I learned from BeOS was what not to do (endian-specific code, C++ ABI, proprietary toolchain, messages on top of byte streams, forced multithreading, etc.). I still don't understand where the responsiveness came from.
The multi-threading was probably part of the reason. SMP was quickly becoming more and more common place with even Power Macs having a second processor despite Mac OS 9 and it’s predecessors not supporting SMP (applications had to specifically be written to support it).
Likewise on Intel, you had Windows 9x that didn’t support SMP so you were stuck with the NT line if you wanted multi processor support. I recall NT4 being pretty poor for desktop usage due to patchy driver support but Windows 2000 was amazing. Linux and BSD weren’t really desktop ready either.
So BeOS was basically a workstation class operating system designed to be running on lower footprint hardware (the comparison someone else made to Amiga is pretty apt).
As much as I typically hate Windows and prefer Linux and FreeBSD, Windows 2000 was pretty solid release and particularly when you place it in its era where only BeOS really competed with it in terms of performance and stability on commodity hardware.
> I still don't understand where the responsiveness came from.
personally, I thought it was because it was macos, which was riddled with a lot of legacy code, for example 68k emulation in the innards of the kernel, and tacked on cooperative/preemptive multitasking.
I remember playing with a leaked version of Copeland/Gershwin. It felt way more modern in a different way than BeOS, but I can't remember how complete it was.
> I still don't understand where the responsiveness came from.
You can get a lot of snappiness from prioritizing GUI processes and lowering the priority of everything else. In modern machines, with lots of cores, you can also try to increase affinity of different interactive processes with a set of cores so that they avoid migrating and losing caches. Finally, Apple is placing low-priority process in the efficient cores and leaving interactive code on the performance ones.
I had a friend who was very excited to show me BeOS running on a PowerMac 7600 back when we were in high school (1997 probably?)
I thought it was interesting, including the UI being so different from what I’d seen otherwise (Mac OS 8 or Win3.1/95) but didn’t appreciate what it really was or why it would be appealing.
In retrospect I’m of two minds. Competition is great, and having more than a duopoly of operating systems wold really put the pressure on to keep competitors honest. That said, I remember the dark days of the single short aisle of Mac programs at EggHead software, and how frustrating it can be to have few options. Really the only thing that makes ChromeOS possible is web applications
> This version has some special software that allows it install without repartitioning a Windows PC’s hard drive
If memory serves me right, it was still possible to install BeOS from that booted OS image file onto whole partition, at the same time wiping existing Windows installation.
Also I've run around Be FAQ site:
> Will the BeOS run on the "iTanium" processor?
Was that CPU really named with "i-thing" fashion back then?
Heck no. BeOS was a tech demo that looked great because you couldn’t do anything complex with it. The list of missing features is endless. The longevity of the platform is unknowable. And there is no reason to think the injection of BeOS would have abated the steady decline of Apple’s relevance.
NeXT gave Apple a mature platform which has withstood the test of time, underpinning multiple architecture changes and adaptation to dramatically different devices. And more importantly, it gave Steve Jobs back to Apple, who orchestrated its recovery from a decline into irrelevance.
A decent web browser was the most obvious problem for a long time. Be was different enough that it was hard to port anything that expected a Berkley socket interface, which was all internet applications.
Which at the time would have been super super duper important for the just-hanging-in-there Mac user base. (I was one of them. It was just hanging in there.)
The Mac’s professional niche was printing. Prepress. Big press runs. QuarkXPress. And BeOS could barely print, and had no color management to speak of.
Yeah, DisplayPostscript/DisplayPDF was a very big deal and upgrade with MacOS X right from the start. That’s an underrated advantage that NeXTStep brought to the table.
Only the first server version of macOS X used DisplayPostscript, Apple replaced it with "quartz" . DisplayPDF. so they would not have to pay the license fee to Adobe.
Correct. But it’s important to note that PDF is an evolution of the postscript language; it shares all of the same drawing primitives. In some respect it’s analogous to JavaScript and JSON. (Yes, postscript is a full scripting language.)
It’s unlikely that Apple needed to strip much out of DisplayPostscript in order to rename it DisplayPDF (and then Quartz).
Starting with the basics, I don't think BeOS even supported multiple users, permissions, access control, etc. I did play around with it in the mid-1990's and recall it had none of this. Perhaps this changed in later versions. But NeXT, other Unix OSes, and NT were far ahead in that regard.
Maybe in 1990. As soon as you start networking your systems in an office network, you do. Keeping your personal files separate from the OS/system files is also a good practice in general, to prevent accidental corruption, etc.
BeOS was tremendously snappy on that hardware but I ended up missing applications on the Mac.