Left Behind: A BeOS and Be Incorporated Post-Mortem

Abstract:

Combating the popular notion that BeOS and Be Incorporated failed mostly at the hands of Microsoft, this article will examine other factors that contributed to Be's downfall including a small development team, underfunding, product maturity, events in the wider computer industry, a paucity of third party developers and applications, and timing.

An MP3 version of this article can be found at https://macfolkloreradio.com/2024/07/10/left-behind-be-incorporated-beos-post-mortem.html.

This work is licensed under a

Creative Commons Attribution 4.0 International License.

Introduction

I began a series on the late great BeOS without a plan. Predictably, this lead to some last-minute panic: how would I end the series? Mac Folklore Radio is composed entirely of other people's writing, and nobody had written a postmortem that earned the Mac Folklore Radio¹ seal of approval, for whatever that's worth.

(Hello? Is this Basal Gangster²? Are you busy right now? Oh, okay.)

It's easy to find vague declarations of the BeBox and BeOS failing at the hands of Windows, but I feel there is much more colour and depth to the story, both inside and outside of x86 PCs, Microsoft, and Be itself. I could have just left you dangling. Unfortunately I have decided to write my own postmortem which I will now inflict upon you.

But why me? Though I wasn't a BeOS developer or even an early adopter, I reluctantly gave up access to the family Mac and daily drove BeOS for about a year—arguably its most exciting year—so I was immersed in the community.

Be set out to build a clean, fast, modern, audiovisual-centric operating system for Everyman, and they succeeded. That by itself is to be commended. BeOS sprouted a small but devoted following that lives on in the form of the Haiku project. It felt so fast and responsive that many former users, including myself, still pine for it today, to mention nothing of its unique filesystem and database features.

I have a difficult time calling that a failure. However, there's no denying development of BeOS ceased and that in 2001, Be Incorporated went bankrupt and was acquired by Palm³. But why, and in what environment? Might they have survived under different circumstances?

There are many reasons Be Incorporated ultimately did not survive, centered primarily on its small development team, underfunding, events in the wider computer industry, a paucity of third party developers and applications, and above all else, timing.

Don't Be Late: APIs and Applications

In business as in audiovisual-centric operating systems, timing is everything. Though they achieved their goals, it would be difficult to overstate how long it took the BeOS to become fully baked.

Be Incorporated was founded in 1990⁴, prototype hardware was ready by the start of 1991⁵, and by July 1994 “they were perhaps six months away from having an OS that they were ready to show publicly⁶”.

However, BeOS' application programming interfaces (APIs, the set of libraries and function calls with which developers build applications) underwent many major revisions up until and most especially in Developer Release 9 in May 1997⁷ ⁸. Possibly as a direct consequence of this continuous API turbulence, software was extremely thin on the ground for the following two years⁹ ¹⁰. Prior to DR9, it would have been madness to start building a large application on top of such quicksand. Examining the BeWare listings at be.com, it would have been extremely difficult to use the BeOS full-time on a daily basis¹¹ until sometime in 1999 when the application base grew to a more palatable breadth and depth¹².

With that overview under our belts, consider that nine years is, uhh, a fatally long time in the personal computer industry. A few things happened elsewhere during that period: Windows 3.0, then Windows 95 took over the desktop world; Windows NT was born and began taking over the server market; Linux went from a hobby project, to geek phenomenon, to threatening the existence of enterprise Unix vendors¹³; Commodore¹⁴ and OS/2 faded away; Silicon Graphics shot to stardom and stumbled halfway into oblivion¹⁵; the Copland project began and ended; Apple bought NeXT, and Mac OS X Server 1.0 shipped; the dotcom boom happened, and would soon come to an abrupt end.

A lot can happen in nine years.

May 1997 is approximately five months after the end of Be's acquisition talks with Apple¹⁶. In the previous episode we heard Jean-Louis Gassée assert that Apple choosing NeXTSTEP instead of BeOS was not a technical matter¹⁷. That may be true, but can you imagine trying to sell Apple an operating system with zero¹⁸ significant applications—an operating system whose APIs have not been battle-tested because they never stopped changing, and have been in flux since 1990?

Now imagine it's October 1995 and you're a developer accepting delivery of a brand-new prototype PowerPC BeBox. Not only would you be disappointed to learn of Be discontinuing its hardware a mere 14 months later¹⁹, but you would have to constantly revise your application as the APIs underwent continuous change for the next two years.

Examine the BeOS Preview²⁰ CD²¹ included on the cover of the August 1997 edition of MacUser. Count the number of third-party applications on that disc. There was ... CodeWarrior Lite, and that was about it—a subtle suggestion that if you wanted applications, you should write your own. (A few people did.)

So it is not merely that Be Incorporated succumbed to Microsoft. It took an excruciatingly long time for the small team at Be to hash out the APIs for BeOS and, after the APIs stopped changing, for third-party applications to proliferate. There were many secondary factors, but it is against this eleven year-long backdrop that Be's story played out.

External Factors: Windows

As we heard in the previous episode, Gassée openly steered the media away from framing Be as competing directly with Microsoft²², inviting them to instead highlight Be's focus on geeks and multimedia performance²³. Nevertheless, after Be was left stranded first by AT&T Microelectronics abandoning the Hobbit CPU²⁴, then by Apple withholding technical information about the Power Mac G3²⁵, it is worth reviewing the Microsoft-saturated environment in which x86 BeOS found itself.

We have established that BeOS and its applications did not mature until approximately 1999, the era in which most new desktop PCs ran Windows 95/98. As an evolution of Windows 3.1, Windows 95 offered an impressive degree of backwards compatibility with 16-bit device drivers and DOS applications, but this highly compatible foundation often restricted Windows 95's stability.

On the other hand, Windows NT was intended to be a fresh start²⁶: a modern, portable (as in easily ported to different CPU architectures), highly secure, multi-user, mission-critical platform for the enterprise. Far more ambitious than both BeOS and Windows 95, it demanded proportionately more resources. The first release of NT recommended a 486 with 16MB of RAM²⁷ in an era when home PCs had 4MB or less. Windows NT 3 and 4 would never see, and were never intended to achieve, mass consumer adoption.

There is a tangible shift that is thoroughly hidden amongst the buzzwords surrounding NT, BeOS and other modern OSes. NT, for all the teething trouble, security holes and blue screens of death in early releases, generally freed users from the burden of living in constant fear of crashes that brought down the entire system. During long operations like rendering a movie, modern operating systems obviated the need to put a “do not touch” sign on your computer for fear someone else in the house would be tempted to peek at their e-mail, unintentionally crash the machine, and ruin your day.

BeOS gave users a taste of Windows NT stability on Windows 95-class hardware—an extremely compelling combination. But without applications, providing a taste was all it could do. The world might have been a very different place if BeOS had showed up with even a medium-sized software library pre-Windows 95 launch.

A famous early company slogan was “one processor per person is not enough”. BeOS was intended to be single user, so it never had to carry the complexity of multiuser process and filesystem security, never mind things like Windows group policy-like management. There's a very good reason BeOS on an early Pentium feels orders of magnitude faster than NT 4 on the same hardware—it was much less sophisticated in these respects, so there was dramatically less code to run. Compressed, the free BeOS 5 Personal Edition installer is just 45 megabytes.

In the previous episode's interview, Gassée stated in October 1998, “We hear that there will be a consumer version of Windows 2000[...] It's getting to a point where the agility required for multiple streams of digital media just isn't there. [...] If you tell me that you have a consumer version of a Boeing 747 ... Generality has its advantages, but also its limitations²⁸.”

Be repeatedly pushed the message that other OSes were too big, too bloated, too slow to handle multiple streams of audio and video²⁹, a line that was even parroted in some market analyst reports³⁰ And yet, no multitrack audio or video editing suites ever shipped³¹ on BeOS, save for the two-track³² video editor personalStudio in 2000³³. In other words, multiple streams of audio and video were handled exclusively on other OSes. They were Good Enough with well-established applications. Nobody loved unexpected whole-system crashes, but work got done.

External Factors: Hardware Evolution

However in 1999, just one year later, 450MHz Pentium II systems fell into the “average home computer” price range³⁴ ³⁵—computers that could make running NT at home almost a pleasant experience. NT 5, renamed Windows 2000, was released to manufacturing in December 1999. Windows 2000 offered some desktop comforts new to NT: plug-and-play, Direct3D, DirectSound, and support for more recent DirectX. It also provided SMP support in the workstation edition, though it was artificially limited to two CPUs.

Direct competition with Microsoft or not, none of this could have helped Be's cause. Windows 2000 didn't boot anywhere near as quickly, but its modern foundation made it dramatically more stable than Windows 95, and its general speed on contemporary hardware and responsiveness under load was Good Enough.

Windows XP, the consumer-oriented version of Windows 2000, would bring the modern foundation of NT to Everyman. OS/2 users must have been shaking their heads—in late 2001, XP finally gave the rest of the world a small sample of what OS/2 2.1 had put on their desks in 1993. It took many years, but Windows and consumer PC hardware evolved to the point where BeOS' value proposition was no longer unique.

External Factors: Money

Among the non-Windows factors standing between Be Inc. and world domination, funding is probably second only to timing. Gassée's memoirs “Grateful Geek”³⁶ characterize Be's financial situation as “[continuously jumping] from one ice floe to the next”.

In “The History of Be Incorporated”, Henry Bortman documented that Be was in serious debt³⁷ by late 1995—not surprising for a five year-old company that was yet to ship a product. A well-received demo at Agenda '95 brought in some funding in the spring of 1996.

Gassée stated in 1997, “We had a couple of near-death experiences. I'm not joking when I say that I've seen the whites of the repo man's eyes³⁸.” In 1998, he sounds much more confident³⁹:

Bortman: So, do you have enough funding to keep going for awhile?

Gassée: Oh, yeah. Not only that, but let's remind ourselves: A successful OS is an incredible instrument of wealth creation [...]

The dotcom boom was in full swing, and “if you could say world wide web, people gave you $10 million. If you could say TCP/IP, they gave you $100 million⁴⁰.” And yet even in that environment, armed with a large investment from Intel in 1998⁴¹, Be announced they were refocusing the company on “Internet appliances” in September 1999⁴². This was months before the dotcom bubble imploded; a tacit admission they were looking to do something profitable rather than continue to vie for a spot in the market as the “audiovisual side OS for Windows”, or “the OS for geeks”.

Be went public in July 1999⁴³ in an attempt to raise funds, which went about as well as one might have expected.

External Factors: Microsoft and PC OEMs

Just as there is more to an operating system than a kernel, there was more to the Windows threat than Windows itself. I'll avoid summarizing in order to preserve the very delicate legal language involved. To quote a news report from Mark Berniker dated September 08, 2003⁴⁴:

Be reportedly claimed Microsoft took actions to prevent Be's presence of the BeOS next to Windows on a line of Hitachi Ltd. PCs.

Be claimed that in 1998 its Be Operating System was to be part of Hitachi's pre-installed “dual boot system.” Be says Microsoft was angry with Hitachi's decision and pressured the company with higher prices for its Windows OS. Any price increase would pressure Hitachi's margins on each PC, making it more cost-effective to remove the BeOS.

Also prior to the settlement, Be had said Compaq Computer, now known as Hewlett-Packard, had agreed to develop a new Internet appliance based on the BeOS. In legal documents, Be said the deal fell through after Microsoft Chairman Bill Gates put negotiating pressure on former Compaq CEO Eckhard Pfieffer.

To quote Gassée's “A Crack In The Wall⁴⁵”:

We've been working with a PC OEM that graciously—and bravely—decided to load the BeOS on certain configurations in its product line. However, there's a twist in their definition of “loading.” When the customer takes the machine home and starts it up for the first time, the Microsoft boot manager appears—but the BeOS is nowhere in sight.

It seems the OEM interpreted Microsoft's licensing provisions to mean that the boot manager could not be modified to display non-Microsoft systems. Furthermore, the icon for the BeOS launcher—a program that lets the user shut down Windows and launch the BeOS—doesn't appear on the Windows desktop; again, the license agreement prohibits the display of “unapproved” icons. To boot the “loaded” BeOS, the customer must read the documentation, fish a floppy from the box and finish the installation. Clever.

As an industry insider gently explained to me, Microsoft abides by a very simple principle: No cracks in the wall. Otherwise, water will seep in and sooner or later the masonry will crumble.

Guarding against even the smallest crack is important to Microsoft [...]

To add insult to injury, the double-clickable BeOS launcher no longer worked in Windows ME. Though ME was only a slight evolution of Win98, there were enough architectural changes to break things. Non-expert Windows ME users would also have to create a boot floppy to start BeOS 5.

Be filed an antitrust suit against Microsoft and was awarded $23 million in September 2003, two years after Be went bankrupt. Gassée joked during a Steve Jobs memorial panel in 2011, “And we even got to sue Microsoft [...] so I could put [...] fresh [...] tires on my wheelchair⁴⁶.” Frank Boosman⁴⁷, VP of Developer Relations:

I have personal reasons to be glad this is settled—not that I stand to receive any money, but as a former executive of Be, I was scheduled to be deposed in the case soon. In a situation like that, the best that can happen is that one manages to avoid looking like an idiot—not exactly something to be excited about.

Having said that, it would have been interesting to see the case go to trial. I don't have first-hand information about Be's allegations of interference in its fundraising efforts, and I don't think Be's argument that Microsoft deliberately destroyed the market for Internet appliances would have flown with a judge or jury. But Be made some very specific allegations about anti-competitive behavior in operating system licensing, and I would have been fascinated to see the parties debate that issue in court.

Internal Factors: Hardware Distraction Layer

It would be easy to label the development of Hobbit and PowerPC BeBoxes from 1990-1997 as wasted time. “Be should have targeted x86 from the beginning,” you might say. But recall that Gassée and Steve Sakoman set out to build a multimedia powerhouse—something unimaginable with the commodity hardware of 1990 when many home users still had 8088-class machines.

However, as time progresses, general purpose hardware often subsumes specialized hardware. It may not have been as obvious in 1990 as it is today, but we certainly saw it in the Macintosh world in the years that followed. For a while, it was possible to buy NuBus cards with custom silicon for decompressing JPEGs⁴⁸. These lasted on the market for only a couple of years before mainstream processors became fast enough to do the same work in a reasonable amount of time without extra hardware.

The short-lived and somewhat incompatible $2000USD Apple 8*24 GC graphics card featured an AMD 29000 CPU, which accelerated 32-bit QuickDraw by up to a factor of 30. Just 18 months later, the Quadras shipped with 80% of the speed of the 8*24 GC without the need for a $2000 card or an exotic CPU. The revised built-in video circuitry on the Quadras' motherboards provided nearly the same performance without compatibility quirks.

In July 1993, the AV models of Macintosh Centris and Quadra shipped with basic spoken command recognition thanks to a built-in AT&T 3210 DSP chip. Less than a year later, the first PowerPC Macintoshes shipped with enough horsepower that the DSP was dropped—it was now simpler, not to mention more economical, to accomplish the same task with a general purpose CPU. (It may have helped that AT&T Microelectronics appears to have abandoned the 3210 at this point along with the Hobbit.)

Be followed the same path. The 1990 BeBox prototype housed three Hobbits and two 3210s. These were eclipsed by dual PowerPC 603s in 1995. BeOS-on-x86 arrived in 1998. It took a while, but commodity hardware eventually caught up with Be's vision.

Internal Factors: Hardware Compatibility Rift

The x86 port of BeOS only began in 1997 before shipping as R4 early the following year. The hardware compatibility list was quite modest at launch⁴⁹, growing to a respectable size in 1999⁵⁰.

However, there was no way to have the BeOS experience on a laptop. BeOS predated Intel Centrino and its descendants; every series of laptop had its own bespoke and totally undocumented chipset. Little wonder, then, that Be didn't waste any effort on PCMCIA support, power management, Wi-Fi or even just IrDA, ACPI suspend/resume, and suspend-to-disk. BeOS diehards might joke that its single-digit-number-of-seconds boot time was a form of suspend-and-resume.

Internal Factors: Platform Maturity

One does not have to peek too far beneath the hood of BeOS to see that it was not Microsoft alone that killed it. A few episodes back, former Apple system software engineer Nitin Ganatra stated on the Debug podcast:

Nitin Ganatra: I had played with BeOS a little bit, but it [...] really felt like there was more sizzle⁵¹ than steak [in the BeOS].

Guy English: Yeah—you could texture map video onto a rotating cube, but you couldn't print, really.

Nitin Ganatra: [laughing] Exactly!

What else was missing, and with all the positive press generated by the free BeOS 5 Personal Edition, how bad could the problem be?

Combing through the BeOS Bible reveals the critical MediaKit had already been rewritten in BeOS' lifetime, and the typography engine was next in line for a rewrite as of 1997. The Old Be File System had been rewritten and replaced with journaled BFS, and the graphics subsystem was also undergoing “a massive rewrite”. The OpenGL layer was in the middle of being rewritten⁵² when BeOS 5 shipped. These may have been indicators of platform maturity that failed to impress Apple during acquisition talks.

BeOS' networking was also of some concern. While a preemptively multitasking OS allowed you to run a web and FTP server in the background without much worry, there were other problems. Rumour had it on comp.sys.be that the userland-based⁵³ network stack was written in a week—a quick hack conjured up right before a demo⁵⁴. Certainly on my hardware⁵⁵, the net_server process needed restarting every hour or so. To BeOS' credit, this only involved a dialog box popping up, a button click, and a delay of about five seconds. But (stop me if you've heard this one before) the network stack was also being rewritten. The new, faster, more stable in-kernel BeOS Networking Environment (BONE⁵⁶) was scheduled to ship “soon” as of early 2000, but Be went under before it was released. Some kind engineer leaked an installer for BONE around the time I'd gone back to the shaky world of classic Mac OS full time.

MacUser's BeOS preview knocks the lack of colour management⁵⁷, and interviewers for the BeOS Bible harassed no fewer than three different Be engineers⁵⁸ ⁵⁹ ⁶⁰ about BeOS' desperately primitive printing architecture. By this point it should not be a surprise that it had already been rewritten⁶¹ by 1999. PostScript support had only just arrived around the time of DR9, but only worked with a subset of Type 1 fonts. These deficiencies may have been quietly intentional—evidence of Be's focus on the multimedia future, not the desktop publishing and prepress past. MacUser also points out the absence of a help system like Balloon Help or Apple Guide⁶².

Printer drivers were also scarce. Perhaps symbolic of BeOS' maturity, printing over AppleTalk was eventually implemented, but only over Ethernet on Power Macs (not LocalTalk⁶³); only for printers outside of AppleTalk zones; and only for printers with alphanumeric names. Printer communication would fail altogether if the printer name contained a slash, like LaserWriter 12/600. The BeOS Printing FAQ⁶⁴ suggests renaming the printer to “something more POSIX-friendly”. Unix geeks and programmers are already shouting, “Just escape the slashes in printer names! Just solve the problem!” Like I said: symbolic. There may have been more underlying architectural gore that precluded such simple solutions.

Related to printing, Unicode support was added to the text handling routines around DR9⁶⁵. Do not confuse the presence of Unicode with internationalization and localization support, of which there was none⁶⁶. The BeOS Bible's interview with Hiroshi Lockheimer reveals, “Currently our strings are all embedded in the source code, which is sort of not a good thing. We want to switch over to resource-based strings pretty soon, but we don't have that mechanism right now⁶⁷ [...]” This is a bizarre omission for a company founded by a native of France accompanied by Apple refugees who must have known about the Resource Manager, or at least seen ResEdit in action, when they began designing BeOS.

Though BeOS featured built-in web and FTP servers, there was no LAN file sharing client until R4.1 in early 1999—the `cifsmount` command. This was particularly painful for Macintosh refugees since BeOS only read and wrote to HFS partitions, not HFS+. An AppleShare client would have made file exchange much easier. BeBits lists a free NFSv2 client, but NFS was not something the average home user cared about.

It cannot have helped Be's case during the NeXTSTEP/BeOS bakeoff that the BeOS APIs were still in flux, and that significant applications were nowhere to be seen as a result. Even if Apple had acquired Be, no infusion of cash would have provided instant relief for a codebase where several major components either needed to be rewritten or were still in the process of being rewritten.

Internal Factors: Applications

I classify applications as an internal problem because they were something Be had some direct influence over in the form of stable APIs or more aggressive developer support. In an infamous user group presentation in 1993, David Barnes of IBM addressed the total absence of retail OS/2 applications by stating, “... we have a group in IBM that's going out there and paying⁶⁸ people to write code.” Of course, providing financial incentives for application development wouldn't necessarily have saved BeOS, particularly since money was not abundant (see “External Factors: Money”) and its APIs were in flux until mid-1997. Intentional or not, Be's strategy appears to have been “entice developers with new platform; let applications grow organically”.

There were a few success stories. TuneTracker⁶⁹ reportedly had BeOS worming its way into the radio automation market; Kenny Carruthers' Postmaster⁷⁰ ⁷¹ was a beloved Eudora-like e-mail client; and Gobe Productive, a ClarisWorks 1.0-like package for BeOS shipped in August 1998⁷², though users hoping for Word and Excel file format compatibility would have to wait another 12 months⁷³. A handful of PC games were ported thanks to OpenGL support and the POSIX compatibility layer: Descent, Doom, Quake, Quakeworld, and Abuse⁷⁴.

There were several image editing packages, some sophisticated, some like MacPaint, and even a substantial vector illustration program, Ornament Tool⁷⁵, which shipped in the summer of 2000. But collectively, to borrow a Brewster Kahle quote, none of them caused the sun to come up a different colour⁷⁶.

After years of proclaiming competing operating systems couldn't cope with handling multiple streams of media⁷⁷, ex-NeXTSTEP developers Adamation⁷⁸ ⁷⁹ ⁸⁰ ⁸¹ shipped personalStudio for BeOS⁸², a non-linear video editing package that supported two (...) video tracks, rather like Adobe Premiere 1.0. Unlike early Premiere, personalStudio allowed you to see transitions and effects in realtime without additional hardware, thanks to hand-optimized assembly code. (Final Cut Pro did not gain CPU-driven realtime effects until version 3⁸³ in late 2001.) Like many other products, MediaPede's UltraDV⁸⁴ and MGI's VideoWave⁸⁵ video editing suites were announced but never shipped.

Step back for a moment and examine the timeline we've followed. personalStudio 1.0 shipped in April 2000, two years after the BeOS MediaKit was rewritten, three years after the major API changes in BeOS DR9, five years since prototype PowerPC BeBoxes shipped (personalStudio was Intel-only), and ten years since Be was founded.

It's worth repeating: Be kept claiming the competition did a poor job of handling media, but audio and video had to be handled exclusively on the competition because no packages like personalStudio shipped until months before Be folded.

Turning to professional audio, things initially look much better. Thumbing through the BeOS Bible⁸⁶ and the press releases at be.com⁸⁷, you'll see an impressive series of commitments to port major applications: Emagic's Logic Audio, Steinberg's Nuendo, EMU Systems Audio Production Studio, and BIAS Peak⁸⁸.

There was just one minor problem: none of these products ever shipped.

Amongst the audio software that did ship, there were a few MP3 players⁸⁹, the DJ suite Final Scratch, a SoundEdit-like package, and a few standalone synthesizers and effects⁹⁰. Most of these tools could not synchronize with one another, rendering them useless for beat-based music performance and production.

Maxon's Cinema 4D was also promised for a couple of years. The BeOS Bible dedicated several pages to showing off and walking the reader through Cinema 4D⁹¹. Its engine (with no user interface, just a viewing window) can be seen in the 1998 BeOS demo VHS tape⁹².

However, like the aforementioned audio tools, the relevant press releases were issued in 1998, about a year before Be gave up on BeOS and shifted its focus to Internet appliances. That alone may have been enough to scare off developers. Be's user base was a tiny fraction of a percent of even that of the Macintosh. Be's shift in focus did not bode well for its growth in the desktop market.

“But wait a minute”, shout the people reading the Wikipedia article on Cinema 4D⁹³. “It says right here that Cinema 4D shipped for BeOS!” Read the source a little more closely; it's a press release stating the port is “being finalized ⁹⁴”. The Wayback Machine shows Maxon's website mentioning BeOS once more in 1999, announcing they intended to wait for the next major release of Cinema 4D before shipping the BeOS port⁹⁵. After Be's shift to Internet appliances, Maxon—and every other major developer—never mentioned BeOS again⁹⁶. Even at the peak of BeOS hype in 2000, when the free BeOS 5 Personal Edition was downloaded more than one million times in the month following its release⁹⁷, the appallingly tiny BeOS application base made even OS/2's software library look good⁹⁸.

133 and Be: A Personal BeOS Narrative

Our family has always had at least one computer in the house going back to the Apple IIe. I was introduced to the Macintosh SE—with dual floppies, no hard disk!—in 1987. We progressed through an SE/30, a IIsi, an LC 630, and in 1998, a Beige G3/266.

In the middle (1996) my parents bought a Pentium 133 with a whopping 64MB of RAM. Why a PC? Because Apple was going to die very soon (which was true at the time) and they figured I should know something about Windows 95 if I was ever to be employable. When the year 2000 arrived, the Pentium 133 had already been gathering dust in the closet for a year or so. My sister was in crunch mode, trying to finish her last few semesters of university. We were in some conflict over who got the G3 and when. I was deeply attached to computers—perhaps you've heard my podcast—and I desperately needed one to myself.

We still had the LC 630 in the closet. Though I'm sure you'd love to have one in your collection today, in 2000 I never even considered going back. The regression in speed and software selection, and the loss of capabilities going back just a few years, like MP3 playback, were too great to fathom.

And then it happened. I think I read it on MacCentral, one of my then-daily haunts. The Mac press ceased paying attention to Be after Apple acquired NeXT. So all at once, I learned Be was still alive, they had ported BeOS to x86, and they were giving away BeOS 5 Personal Edition for free. Whoa! It looked like it would run on my motherboard, too, an Intel 430FX-based Asus P55TP4N. (Trainspotter alert.)

Everyone—well, every geek—had heard about Linux and the open source movement, so this move was unusual but not entirely surprising. It was free as in beer; it was not open source. You could download the product yourself or pay $70USD for BeOS Pro on CD-ROM⁹⁹, which added RealPlayer G2, the BInkjet¹⁰⁰ set of printer drivers, an MP3 encoder, the Intel Indeo video codec for those of you old enough to remember that, and a “light” version of Partition Magic to assist with installation.

I extracted the Pentium 133 from the closet and attached an equally unloved Apple Multiple Scan 15AV monitor I'd nicknamed “Fuzzy” for its 0.28mm dot pitch. I remember BeOS booting in exactly 10 seconds. Surely this was a joke or a trick, or just a bootloader that displayed a static screenshot. Windows 95 OSR2 took about 90 seconds to boot on the same hardware. Just to emphasize: it booted so fast that once your BIOS initialized and your monitor caught up when the BeOS bootloader changed video modes, the boot screen was only visible for three seconds. I have a fairly good memory for these things, but all I can remember is the word “BeOS” and some icons in circles.

BeOS made everything feel so immediate and responsive. I never had to worry about interrupting network transfers when the mouse button was held down. When something critical was happening, I didn't have to plan my every move to avoid crashing the machine. The journaling filesystem protected you from directory corruption. All of this on an elderly Pentium 133? This was what Apple had been promising us since 1994, and I longed to have my favourite Macintosh applications in this kind of environment.

I could see why they had waited to make a big splash with the free release: a smattering of mature, useful applications had finally accrued. (The last time I looked in 1998, it had been mostly tech demos and toy apps.) The built-in web browser, NetPositive, had the capabilities of Netscape 2.0 from five years prior, but that was still more than acceptable back in 2000. It crashed every 20 minutes but, very unlike the Mac, you simply re-opened it and kept going. I always had a NetPositive window open with www.bebits.com¹⁰¹ loaded, a running list of new software and updates in the style of VersionTracker¹⁰² and MacUpdate¹⁰³.

Someone had ported the open source word processor AbiWord¹⁰⁴ to BeOS. Maybe I could do my homework in this instead of the built-in SimpleText clone. It looked nice enough but crashed whenever you typed more than about six words. Oh well.

Thankfully other software proved to be more mature. Since I was without access to HFS+ volumes, I FTPed MP3s and text files to the spare Mac in the basement. The Eudora-like Postmaster¹⁰⁵ e-mail client was of great comfort.

Other must-haves included CL-Amp for playing MP3s, a DivX decoder, the VICE Commodore 8-bit emulator suite, SheepShaver (it was originally written for BeOS), Basilisk II, and ICBM, a client for the ICQ instant messaging service.

When I graduated from high school, my parents offered to buy me a computer as a graduation gift. I was too cool to ask for an iMac, and at nearly twice the price without a monitor, asking for a Power Mac G4 felt greedy. I pondered going for a dual Celeron ABit BP6¹⁰⁶ because having two processors for BeOS seemed cool—even if they were horrifically slow—but a friend convinced me to go with an Athlon 750 instead. I've said this several times through out this series, but BeOS felt so fast on a Pentium 133 that stepping up to an Athlon 750 was a disappointment because it didn't feel any faster!

When I got a summer job, I stupidly ordered the boxed copy of GoBe Productive instead of the electronic download, and it took four months to arrive. The first public version of the BeShare¹⁰⁷ peer-to-peer file sharing client arrived that summer. Yes, BeOS users had their own exclusive piracy platform, mostly sharing—when I looked—MP3s and MPEG-1 music videos. Hungry for applications, I was hoping to see some being traded, but none ever flew past. It felt good to have a new machine with a modern OS that was completely Microsoft-free.

In the fall I began my first semester of university plus a part time job on campus helping an extremely Mac-centric department. Unfortunately Gobe Productive still hadn't arrived in the mail, and I had a class that required (I am not making this up) some 16-bit Windows 2.0 software. I also needed a “real” word processor, not just a text editor, for the few assignments I bothered to complete. I reluctantly installed Windows 2000 on a second hard disk.

As previously discussed, I didn't want to admit it but even a relatively affordable Athlon system could run NT 5—aka Windows 2000—at a tolerable speed. It wasn't anywhere near as fun or responsive as the BeOS, but it was Good Enough, and it had none of the fragility and choppy multitasking of classic Mac OS. I felt dirty running Windows, but the stability had me hooked. I found myself using Windows more and more because I needed access to more applications, occasionally rebooting back into BeOS to cleanse myself.

In late September I happened by Be's website while procrastinating. Front and center was a special feature, “[An] Open Letter¹⁰⁸ From Steve Sakoman, Be. Inc. COO”—already a bad sign from a company undergoing a major shift in focus. I clicked. I saw “BeOS” in the first line and my heart sank. BeOS was probably dying (what can I say, I'm an optimist) and Be would probably die with it unless they somehow made hundreds of millions of dollars with Internet appliances. You certainly didn't see any open letters from Microsoft about Windows.

By December 2000, the hype generated by the free distribution of BeOS 5 Personal Edition seemed to have fizzled out altogether. I was a little disillusioned. Where was the multitrack audio recording and editing software for which Be repeatedly proclaimed it had built a great foundation? Would marketing people lie? There was still only a two-track DV editing package. No standalone professional audio products ever appeared. It may have been my imagination, but the flow of new software on BeBits seemed to have slowed down quite significantly.

A friend snagged a copy of the Mac OS X Public Beta which we eagerly installed on his tray-loading iMac G3/333. We were a little worried. The beta was very stable but agonizingly slow. Applications took 20, 30, 40 seconds to launch. We tried updating the prebinding. No difference. Maybe the release version would be faster. But what if it wasn't?

In January 2001, a few days after the Macworld San Francisco Stevenote, I'd had more than enough of Windows and used my earnings to buy a G4/466 Digital Audio with my academic discount. Going back to Mac OS felt like coming home. It had its flaws but the comfort and familiarity made them melt away, coupled with the overwhelming relief of having access to audio toys and other applications again.

I began checking out leaked Mac OS X betas distributed on IRC. They worried me just as much as the Public Beta. Performance had greatly improved by percentage but not perception. The worst case scenario, Carbonized Internet Explorer 5, was down to a launch time of 15 seconds. This was worlds away from BeOS, and I felt the future that I'd held in my hands slipping away.

Thankfully 10.1 was significantly faster, and by 10.2 there were enough native OS X applications that I could use it full-time. The dramatic speed improvements in 10.3 and 10.4 began to make it feel more BeOS-like, though the overall experience is still today, as one forum commenter pointed out years ago, clearly designed to be slow and beautiful.

Into the Great Beyond

Maybe Be Incorporated was a bit too much like mid-1990s Apple: great talent driven by bold visions and poor engineering management. Be spent five years without shipping a product, an additional two years before shipping an OS with (mostly) mature APIs, and on top of that, users waited yet another two years for applications to grow. Ironically, Gassée spent six years at Hewlett-Packard where, as Steve Jobs outlined at CAUSE 1998¹⁰⁹:

... the goal used to be to make the best computers in the world. [...] Goal two we got from Hewlett-Packard¹¹⁰, actually, which was “we have to make a profit”, because if we don't make a profit, we can't do goal one!

The Hewlett-Packard aphorism that more companies tended to die from indigestion than starvation¹¹¹ definitely did not apply to Be. The only surprising thing about Be's demise is that they managed to raise as much funding as they did. Nine years between starting a new computer platform and reaping the benefits of its first crop of applications is simply too long to overcome the odds of starving to death. Ten years elapsed between Windows 1.0 and Windows 95, but Microsoft had other highly lucrative products to rely on, and 100,000 employees¹¹² to Be's 100¹¹³.

Had Apple acquired Be, the money wouldn't have helped the missing and half-finished parts of the BeOS codebase. Advances in hardware eventually brought Be's modern multimedia OS vision into wide distribution in the form of Windows 2000 and XP running on inexpensive consumer PCs.

If only the narrow window when Be's application base began to mature (1999-2000) had come a few years earlier. Maybe BeOS would have had a chance to catch up to graphical desktop Linux in popularity if it had been open-sourced. Even Linux advocates had to admit BeOS made ancient hardware feel fast, a trick that X11 and KDE or GNOME could not replicate¹¹⁴.

24 years after I first downloaded BeOS 5 Personal Edition, no corporation seems to care about improving the desktop computer experience. Today's neglected operating systems have deteriorated into zombies, passive host bodies for Docker containers and multiple instances of Google Chrome under the guise of Electron applications. The browser has reinvented desktop applications—poorly—and now we have to worry about inter-browser-tab memory protection and process separation. User interface design consistency, responsiveness, and speed/space optimization have been thrown out the window. As the reliability, simplicity, and performance of macOS continue to rot post-iPhone release, I begin to wonder if we should do with macOS Sonoma what we did with classic Mac OS: throw it out and rehost the APIs on top of BeOS R5.

Andy Hertzfeld¹¹⁵ asked the million dollar question at the Churchill Club Steve Jobs Memorial event in 2011:

Andy Hertzfeld: I can't help but want to ask you, Apple almost bought Be, your company, instead of NeXT. How do you feel about that?

[everyone laughs]

Pam Kerwin: Let's don't go there!

Jean-Louis Gassée: Thank God that didn't happen! You know, ... because I hated—I have to say it—I hated Apple's management. I mean seriously, the discussions—

Andy Hertzfeld: So did Steve!

Jean-Louis Gassée: Yeah, exactly.

Bill Atkinson: He got rid of 'em.

Jean-Louis Gassée: I couldn't picture myself in there, so when we lost the deal it stung, as it should [...] You know, thank God they didn't buy Be, because life would be a different place.

There was a lot missing in BeOS, most of all timeliness and applications, but the free distribution of BeOS R5 reached over a million people¹¹⁶ and it is very fondly remembered. Rand Miller, co-creator of the 1993 hit CD-ROM game Myst said in a 2020 interview¹¹⁷ ¹¹⁸ ¹¹⁹:

I think that probably only 50% of the people who played [Myst] even made it off the Island of Myst [the starting point]. But for some reason, that didn't dampen enthusiasm. I think because that first island was so intriguing and you got enough of the story to kind of be tantalized [...], it left a great taste in people's psyche [...]”.

One could draw a parallel between between these entranced players who never made it off the visually sumptuous Myst Island, thus missing out on 90% of the game, and BeOS with its beautiful foundation and dearth of applications. Despite its ultimately going bankrupt, BeOS users fell in love with the company's backstory, its mission, and the positive difference it made in their lives. And that is what I miss most of all about Be and the computer industry at large that quietly disappeared during the 2010s—the focus on making people's lives better rather than maximizing profits, and doing so in a law-abiding way by solving problems that needed to be solved; empowering the individual so that, in their own small, quiet way they, too, could make the world a better place.

Revision History

20-Nov-2024: Fixed typos and removed redundant paragraph

10-Jul-2024: Podcast and HTML version published

05-Jun-2024: Initial version (not published)

About this document ...

Left Behind: A BeOS and Be Incorporated Post-Mortem

This document was generated using the LaTeX2HTML translator Version 2024.2 (Released October 21, 2024)

The command line arguments were:
latex2html –split 0 index.tex

The translation was initiated on 2024-11-20

Footnotes

... Radio ¹: https://macfolkloreradio.com/
... Gangster ²: http://basalgangster.macgui.com/RetroMacComputing/The_Long_View/The_Long_View.html
... Palm ³: “Be Incorporated [...], the creator of the BeIA and BeOS operating systems, today announced it has entered into a definitive agreement to sell its intellectual property and technology assets to Palm, Inc [...]. The purchase price is $11 million, to be paid in common stock of Palm [...]” https://web.archive.org/web/20010821024413/http://www.be.com/press/pressreleases/01-08-16_assetsale.html
... 1990 ⁴: “Gassée's last day at Apple was September 30, 1990. On October 1, he and Steve Sakoman took a trip to Fry's Electronics.” https://archive.org/details/the-beos-bible/page/30/mode/1up
... 1991 ⁵: “By the end of the [1990], he had built the first Be prototype.” https://archive.org/details/the-beos-bible/page/30/mode/1up
... publicly ⁶: https://archive.org/details/the-beos-bible/page/33/mode/1up
... 1997 ⁷: “By now you're probably starting to absorb the long list of things that have changed for DR9.” https://www.haiku-os.org/legacy-docs/benewsletter/Issue2-20.html,
... ⁸: https://www.haiku-os.org/legacy-docs/benewsletter/Issue2-17.html
... years ⁹: Much of the software available for BeOS one year post-DR9 was still—I'm sorry—toy apps, and Unix command line utilities recompiled for the BeOS POSIX layer. https://web.archive.org/web/19980520212151/http://www.be.com/beware/Available_List.html
... ¹⁰: It took one year after DR9 for a word processor and spreadsheet package to appear. https://web.archive.org/web/19980520224644/http://www.be.com/beware/Productivity/Be%20Basics.html
... basis ¹¹: https://web.archive.org/web/19970525184837/http://www.be.com/beware/
... depth ¹²: https://web.archive.org/web/20000815060354/http://www.be.com/software/
... vendors ¹³: “It is certainly the case that the rise of Linux helped put pressure on Sun Microsystems.” http://esr.ibiblio.org/?p=6279
... Commodore ¹⁴: https://www.commodore.ca/commodore-history/commodore-bankruptcy-associated-press-1994/
... oblivion ¹⁵: “Silicon Graphics experienced mounting losses in fiscal 1998 [...] lost $55.5 million on revenues of $768 million [...] More than half of Silicon Graphics' sales came from shrinking markets such as Unix workstations and supercomputers [...] https://www.company-histories.com/SGI-Company-History.html
... Apple ¹⁶: Apple and NeXT began discussions in November 1996 [...] also in negotiation with Be [...] https://medium.com/chmcore/next-steve-jobs-dot-com-ipo-that-never-happened-3cad4a45f3e9
... matter ¹⁷: “Clearly, the technical choice was not the issue.” https://archive.org/details/the-beos-bible/page/40/mode/1up
... zero ¹⁸: https://web.archive.org/web/19961029235127/http://www.be.com/beware/index.html
... later ¹⁹: “The 66MHz BeBox began production in 10/95 [...] and was manufactured until 12/96. The 133MHz BeBox began shipping in 10/96, and ended when Be announced that we would discontinue hardware production, on 1/30/97.” http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/support/qandas/faqs/faq-0261.html
... Preview ²⁰: https://web.archive.org/web/19970606090956/http://www.be.com/products/beos_tour/index.html
... CD ²¹: https://archive.org/details/be-os-preview-release
... Microsoft ²²: “Microsoft is the de facto standard. It's the weather. And my view is, don't fight the weather, it doesn't work. [...] Let's say [...] I'm selling audio and video systems for vehicles and you just bought this new Suburban [...]. You come to my shop and I say, `Where did you buy this piece of...?' If I start criticizing your vehicle [...] am I going to make friends with you? Are you going to listen to me?” https://archive.org/details/the-beos-bible/page/41/mode/1up
... performance ²³: “But audio and video applications, which demand more bandwidth, are more naturally coherent with Intel's margin structure. We help create demand.” https://archive.org/details/the-beos-bible/page/41/mode/1up
... CPU ²⁴: Hermann Hauser: “And what [Larry Tesler] argued—and of course he turned out to be completely right—was that he did not trust a $2 billion electronics subsidiary of AT&T, AT&T microelectronics. He did not trust them to support that microprocessor.” https://archive.computerhistory.org/resources/access/text/2015/04/102739951-05-01-acc.pdf#page=38
... G3 ²⁵: http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/support/qandas/faqs/faq-0408.html
... start ²⁶: “Cutler had three main aims for his operating system: [...] Portability [...], Reliability [...], Personality[...].” Zachary, G. Pascal (1994). Showstopper: The breakneck race to create Windows NT and the next generation at Microsoft (pp. 48-49). The Free Press.
... RAM ²⁷: The back of the Windows NT 3.1 box lists 12MB RAM and 75MB free disk space as the `recommended minimum' for x86; 16MB RAM/92MB free disk space for `RISC' systems.” https://virtuallyfun.com/2009/08/19/networking-with-windows-nt-3-1-under-emulation/
... limitations ²⁸: https://archive.org/details/the-beos-bible/page/44/mode/1up
... video ²⁹: The BeOS Bible repeated this message: “No matter how fast our processors and video cards become, we still need an operating system capable of handling multiple simultaneous data streams without clogging up.” https://archive.org/details/the-beos-bible/page/6/mode/1up
... reports ³⁰: “The company was originally formed around a desktop computer operating system that was designed to augment [...] Windows and Apple?s Mac OS, particularly for audio- and video-intensive applications that aren't easily handled by either system.” https://plotkin.com/cnbcs141/
... shipped ³¹: BeOS found a home in two pieces of pro and semi-pro gear: the Edirol DV-7 video editing appliance, and the Otari RADAR 24 multitrack hard disk recording system. I have excluded these because they were combination hardware/software solutions that cost $5,000 and $25,000USD respectively, far out of the reach of home users.
... two-track ³²: In the context of non-linear editing, "two tracks" != "multi-track" IMHO.
... 2000 ³³: https://web.archive.org/web/20000309072135/http://www.adamation.com/Products/personalStudio/
... range ³⁴: https://web.archive.org/web/19991127182900/http://www.dell.com/us/en/dhs/products/series_dimen_desktops.htm
... ³⁵: https://archive.org/details/pc-computing-magazine-v12i3/page/n2/mode/1up
... Geek”³⁶: https://gratefulgeek.co/book/
... debt ³⁷: https://archive.org/details/the-beos-bible/page/35/mode/1up
... eyes ³⁸: https://archive.org/details/the-beos-bible/page/36/mode/1up
... confident ³⁹: https://archive.org/details/the-beos-bible/page/45/mode/1up
... million ⁴⁰: Unfortunately I've forgotten the source of this quote.
... 1998 ⁴¹: “[...] Be Incorporated will announce that chip giant Intel has taken an equity stake in the company [...]” https://web.archive.org/web/20010518152653/http://news.cnet.com/news/0,10000,0-1003-200-335301,00.html
... 1999 ⁴²: https://web.archive.org/web/20010209052940/http://www.be.com/press/pressreleases/99-08-31_ipad.html
... 1999 ⁴³: https://web.archive.org/web/20170212124701/https://plotkin.com/cnbcs141/
... 2003 ⁴⁴: https://web.archive.org/web/20131109045719/http://www.internetnews.com/ent-news/print.php/3073811/
... Wall ⁴⁵: “Be Newsletter 4-8 (February 1999): A Crack in the Wall“: https://www.haiku-os.org/legacy-docs/benewsletter/Issue4-8.html#Gassee4-8
... wheelchair ⁴⁶: https://youtu.be/N2C2oCsrqcM?t=3671
... Boosman ⁴⁷: https://web.archive.org/web/20040224003721/www.boosman.com/blog/archives/000931.html
... JPEGs ⁴⁸: https://ieeexplore.ieee.org/document/128824
... launch ⁴⁹: https://web.archive.org/web/19980520211743/http://www.be.com/products/beosreadylist.html
... 1999 ⁵⁰: http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/support/guides/beosreadylist_intel.html#chipsets
... sizzle ⁵¹: 58m54s: https://www.imore.com/debug-39-nitin-ganatra-episode-i-system-7-carbon
... rewritten ⁵²: “[...] there is no hardware acceleration for OpenGL in BeOS 5 [...] One reason for this is that we were in the middle of a major re-write of OpenGL [...]” https://www.haiku-os.org/legacy-docs/benewsletter/Issue5-13.html
... userland-based ⁵³: “Many of BeOS's critical system services are handled by [userspace] servers running in the background. The net_server, for instance, ...” https://archive.org/details/the-beos-bible/page/328/mode/1up
... demo ⁵⁴: I read this back in 1997 but have been unable to locate the original Usenet post.
... hardware ⁵⁵: An ASUS K7V motherboard with a 3Com 3C905b NIC
... (BONE ⁵⁶: https://www.haiku-os.org/legacy-docs/benewsletter/Issue5-5.html
... management ⁵⁷: “The BeOS' lack of color management is going to make winning over prepress professionals hard.” https://archive.org/details/MacUser9701January1997/page/n74/mode/1up
... engineers ⁵⁸: Bortman: “When things like printing are fully implemented, [...]” https://archive.org/details/the-beos-bible/page/396/mode/1up
... ⁵⁹: Bortman: “Printing. It's not exactly BeOS' strong suit.” https://archive.org/details/the-beos-bible/page/493/mode/1up
... ⁶⁰: Bortman: “I have to ask this. What about printing?” https://archive.org/details/the-beos-bible/page/740/mode/1up
... rewritten ⁶¹: “1999: The BeOS has undergone a printing architecture rewrite.” http://asleson.org/public/mirrors/www-classic.be.com/support/breakingnews/html/whatprints.html
... Guide ⁶²: “No help system [...and] none on the drawing board. https://archive.org/details/MacUser9701January1997/page/n75/mode/1up
... LocalTalk ⁶³: https://archive.org/details/the-beos-bible/page/458/mode/1up
... FAQ ⁶⁴: http://asleson.org/public/mirrors/www-classic.be.com/support/breakingnews/html/printing.html
... DR9 ⁶⁵: “You may have heard that DR9 adopts the Unicode Standard for encoding characters [...]”. http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/aboutbe/benewsletter/Issue75.html#Insight2
... none ⁶⁶: As of April 1999, localization was still a third-party product opportunity. http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/beware/Development/Blithe.html
... now ⁶⁷: BeOS Bible interview with Hirsoshi Lockheimer: https://archive.org/details/the-beos-bible/page/575/mode/1up
... paying ⁶⁸: At 1h10m30s in https://archive.org/details/OS2NTShootout or 1h50m38s in https://youtu.be/-DAojx2Hgec?t=6638
... TuneTracker ⁶⁹: http://www.tunetrackersystems.com/ttbasic.html
... Postmaster ⁷⁰: https://web.archive.org/web/20001001094610/http://www.bedepot.com/products/KennyC/Postmaster.asp,
... ⁷¹: https://web.archive.org/web/20020204105617/http://kennyc.com/postmaster/index.html
... 1998 ⁷²: https://web.archive.org/web/19980625062410/http://www.gobe.com/index.html
... months ⁷³: https://web.archive.org/web/20000301120224/http://www.gobe.com/2-0.html
... Abuse ⁷⁴: https://web.archive.org/web/20000816213409/http://www.bebits.com/browse/89
... Tool ⁷⁵: https://web.archive.org/web/20010202073500/http://ornamente.de/
... colour ⁷⁶: At 5m15s: https://archive.org/details/nerdtv004
... media ⁷⁷: “...the agility required for multiple streams of digital media just isn't there...” https://archive.org/details/the-beos-bible/page/42/mode/1up
... Adamation ⁷⁸: https://computerhistory.org/blog/meet-the-adams-brothers/
... ⁷⁹: Computer History Museum - Oral History of Stephan Adams and William Adams https://archive.computerhistory.org/resources/access/text/2021/08/102792110-05-01-acc.pdf
... ⁸⁰: CHM Oral History of Stephan and William Adams Part 1 of 2 https://youtu.be/WBkvaNDPuxA
... ⁸¹: CHM Oral History of Stephan and William Adams Part 2 of 2 https://youtu.be/Xk8DdHdc0Ao
... BeOS ⁸²: “personalStudio is Adamation's real-time, full-screen, non-linear video editing and compositing tool for digital media production.” https://web.archive.org/web/20000309072135/http://www.adamation.com/Products/personalStudio/
... 3 ⁸³: “[...] the new G4 real-time effects in Final Cut Pro 3 deliver the [...]” https://www.apple.com/newsroom/2001/12/04Apple-Announces-Final-Cut-Pro-3/
... UltraDV ⁸⁴: “Mediapede is [...] dedicated to creating [...] tools for the BeOS. [...] Regretfully, development of UltraDV has been suspended [...]” https://web.archive.org/web/20000823223458/http://www.mediapede.com/index.html
... VideoWave ⁸⁵: https://web.archive.org/web/20001202062200/http://www.mgisoft.com/video/videowave4/index.asp
... Bible ⁸⁶: “Coming Soon: At this writing [1998], a number of [...] had just been announced for BeOS. [...] These products should be on the scene by mid-1999.” https://archive.org/details/the-beos-bible/page/768/mode/1up
... be.com ⁸⁷: be.com Press Release Archive https://web.archive.org/web/20010123220300/http://www.be.com/press/pressreleases/pressreleases2.html
... Peak ⁸⁸: “[...] BIAS today announces plans to port its highly-acclaimed Peak^TMdigital audio editor to BeOS[...]” https://web.archive.org/web/20010308231621fw_/http://www.bias-inc.com/press/pr990809_1.html
... players ⁸⁹: https://web.archive.org/web/20000815094823/http://www.bebits.com/browse/108
... effects ⁹⁰: https://web.archive.org/web/20000816212210/http://www.bebits.com/browse/195
... 4D ⁹¹: https://archive.org/details/the-beos-bible/page/718/mode/1up
... tape ⁹²: https://youtu.be/BsVydyC8ZGQ?t=503
... 4D ⁹³: https://en.wikipedia.org/wiki/Cinema_4D
... finalized ⁹⁴: http://testou.free.fr/www.beatjapan.org/mirror/www.be.com/aboutbe/pressreleases/98-11-10_maxon.html
... port ⁹⁵: "BeOS portation of CINEMA 4D comes after AMBER": https://web.archive.org/web/20000819083909fw_/http://www.maxon.de/pages/news/news_e.html
... again ⁹⁶: Maxon's archived press releases do not indicate any BeOS products ever shipped. https://web.archive.org/web/20001209213200/http://maxon.de/usa/index.html; https://web.archive.org/web/20010111043700/http://maxon.de/index_e.html
... release ⁹⁷: “Be Goes Platinum with BeOS 5: Over One Million Copies of BeOS 5 Personal Edition Downloaded in First Month” https://web.archive.org/web/20010110031900/http://www.be.com/press/pressreleases/00-05-09_platinum.html
... good ⁹⁸: There was at least one multi-track (destructive, not DAW-like) sound editor for OS/2. http://www.os2voice.org/warpcast/1998-04/352CB50F.htm; https://web.archive.org/web/19980211130157fw_/http://www.ceressoft.com/sstudio/index.html
... CD-ROM ⁹⁹: https://web.archive.org/web/20001026021949/http://www.gobe.com/beos/beosinfo.html
... BInkjet ¹⁰⁰: http://www.clarkwood.com/old/binkjet/
... www.bebits.com ¹⁰¹: https://web.archive.org/web/20000815094940/http://www.bebits.com/
... VersionTracker ¹⁰²: https://web.archive.org/web/19980130182555/http://www.versiontracker.com/
... MacUpdate ¹⁰³: https://web.archive.org/web/20000303232252/www.macupdate.com
... AbiWord ¹⁰⁴: https://web.archive.org/web/20000816214855/http://www.bebits.com/app/356
... Postmaster ¹⁰⁵: https://web.archive.org/web/20020204105617/http://kennyc.com/postmaster/index.html
... BP6 ¹⁰⁶: https://www.anandtech.com/show/343/1
... BeShare ¹⁰⁷: https://web.archive.org/web/20000815095159/http://www.bebits.com/app/1330
... Letter ¹⁰⁸: https://web.archive.org/web/20000815060544/http://www.be.com/press/sakoman_letter.html
... 1998 ¹⁰⁹: https://youtu.be/e-B9LikzxyM?t=1790
... Hewlett-Packard ¹¹⁰: https://fs.blog/the-hp-way-david-packard/
... starvation ¹¹¹: Packard, David. The HP Way (Harper Business, New York, 1995)
... employees ¹¹²: Approximately 100,000 employees at the time of Windows NT 3.1's release in 1993: https://stockanalysis.com/stocks/msft/employees/. Of course, not all of these were software engineering working on Windows, but the point is that Microsoft had a far larger pool of engineers to draw upon.
... 100 ¹¹³: Be laid off 28 employees “[representing] approximately 33% of the company's existing workforce.”. https://web.archive.org/web/20010807065004/http://www.be.com/press/pressreleases/01-07-31_streamlining.html
... replicate ¹¹⁴: https://www.osnews.com/story/7324/the-fast-food-syndrome-the-linux-platform-is-getting-fat/
... Hertzfeld ¹¹⁵: https://youtu.be/N2C2oCsrqcM?t=3624
... people ¹¹⁶: Though not all Be employees enjoyed the adventure: https://archive.org/details/the-cotton-squares-collection/The+Cotton+Squares+Collection+-+Virtual+Void+and+other+BeOS+Songs+and+more/The+Cotton+Squares+Collection/Cotton+Squares+and+Baron+Arnold+tunes/Pura+Como+el+Agua.mp3
... interview ¹¹⁷: https://cdn.arstechnica.net/wp-content/uploads/2020/01/Myst_Transcript.txt
... ¹¹⁸: https://arstechnica.com/video/watch/war-stories-myst
... ¹¹⁹: https://arstechnica.com/gaming/2020/02/an-extended-interview-with-atrus-himself-myst-creator-rand-miller/