Gassée: Thank God Apple chose NeXT over my BeOS
9to5mac.com
9to5mac.com
But Gassée fumbled it when courting Apple [1], and then sold it to Palm, who promptly killed it. It has taken years for a dedicated group of followers to rebuild it from scratch, and by the time they finished doing that, computer hardware had advanced so much that there was very little demand for a high-performance operating system in the desktop market.
So, I guess he can claim to be glad that his baby died a lonely and ultimately unremarkable death thanks largely to his own mishandling of it, but it sounds like sour grapes to me.
[1]: I remember reading an article a while back about the meeting between Amelio, Gassée, and Jobs. Reportedly, Gassée thought BeOS was a shoe-in, despite asking an astronomical figure for the sale, so he showed up to the meeting completely unprepared. Steve, in typical Steve fashion, showed up fully prepared and blew Gassée right out of the water. I've spent fifteen minutes searching for that article, and I can't find it [2] -- and it might be apocryphal anyway, according to a post on Slashdot (http://apple.slashdot.org/comments.pl?sid=129504&cid=108...). Still, Gassée and Jobs have had an uneasy relationship in the past (http://folklore.org/StoryView.py?project=Macintosh&story...).
[2]: Woo, found it! http://macspeedzone.com/archive/art/con/be.shtml
edit: aside from that, it was great to see and hear Bill Atkinson, and Gassée had some funny quotes.
This is either sour grapes because the deal fell through or else the deal fell through because he let personal issues get in the way of business. I lean strongly towards the latter, given Gassée history and the unreasonable price tag he put on BeOS. http://en.wikipedia.org/wiki/Jean-Louis_Gass%C3%A9e#1980s:_A...
During the BeOS negotiations rumor had it that the sticking points were - aside from money -largely cultural and managerial. None of the former Apple people at BeOS wanted anything to do with Apple management and they were negotiating for things like not having to be in Apple's org chart. This was NOT a recipe for success.
Too many times i was also fooled bymyself in thinking negotiations, be it companies or even buying a car, followed a logical progression.
As was NeXTSTEP. The difference being it was, by that point, a mature OS and framework set with a wide variety of actual deployed applications, cross-platform support, tooling. BeOS, in comparison, was a clever and incomplete tech demo. And Apple had no shortage of those (Copland, Taligent, SOM, OpenDoc, PowerTalk, etc, etc, etc). Buying another one would have been catastrophic.
And had been used in important roles in industry, particularly the financial industry.
That was kinda the problem. BeOS was built around the 1990s mentality that PC hardware was quite limited, and a light fast OS was more important than security, robustness, etc. It ran like greased lightning, but really wasn't a whole lot more advanced than other contemporary consumer OSes.
Next, on the other hand, was always intended to be a workstation OS (with 'enterprise frameworks') and was a much better fit for the hardware of the 2000s. With MS moving to the NT line, it would have been a huge mistake for Apple to adopt a single-user OS like Be.
Back in the day, I actually used both the NeXTcube and a BeBox (although neither as a primary environment)
The NeXT environment enamored everybody who was introduced to it... for a day or two. Them almost everybody moved away from it because it was so... damn.. slow. The base cube only had 8MB of RAM (a decent amount in those days!) which wasn't nearly enough for the big heavy frameworks it used. NeXT had to retrofit all of the cubes with 40MB SCSI drives just for swap. The display postscript system was elegant (it provided a common display engine for both the screen and the NeXT-brand laser printer, for instance) but ate up precious cycles of the 68030. The magnetic-optical drive that supposedly was the future of personal storage had horrible performance (although the drives mercifully gave out after a few years anyway)
In short, the NeXTcube's we had were the machines everybody loved but few wanted to actually use.
The BeBox by contrast was nothing if not zippy! There were too many missing bits for it to really be my main environment, but you really felt that you were using the future of operating systems. After years of using UIs that always had at least some lag the BeBox's instantaneous response was almost a shocking experience.
When Apple was quite publicly looking for an OS to buy I was certainly rooting for Be. It seemed to fit their needs perfectly: it already ran on PowerPC and its responsiveness would have been great for the multimedia creators that were Apple's bread-and-butter. When they went with NeXT instead I thought it was just another of Apple's classic missteps. I couldn't see why they were hitching their wagon to an OS that was last decade's news.
Clearly I was wrong, even if you ignore the Steve Jobs factor. The hardware advances underway quickly made the amazing responsiveness of BeOS moot; now every OS feels like the snappy BeBox did in 1995. As memory sizes grew the NeXTstep frameworks came into their own (probably if the NeXTcube could have had 1GB of RAM it would have been fine!)
To put it simply, Be built an OS in 1996 that was perfect for 1996. NeXT build an OS in 1989 that was suited for the 2000s.
I don't mean that last bit as a pure complement, by the way. The "too far ahead of its time" is a cliche about failed tech products. Building something that only works with hardware 10+ years in the future is hardly a recipe for success. It's only through the accidents of history that NeXTstep got its chance to shine years after its original failure. 99% of products in that position only get to be footnotes.
That 68030/optical disk/40MB swap drive era didn't last very long.
NeXTSTEP was quite nice with a 16MB 25MHz 68040, and a decent hard drive, as seen in the NeXTStation from 1990. Even the price was competitive if you didn't mind the 2-bit greyscale monitor.
NeXTSTEP 3.3, running in color on a Pentium Pro (or better), or even better a 1994 HP PA-RISC workstation, was SWEET.
And rightly so. But given that the hardware was built (at great expense!) specifically to run that OS, I think it's fair to consider their performance together.
Certainly the gap between "the hardware the OS needs" and "what is readily available" shrunk as time went on. The slabs with more RAM were substantially better than the cubes had been. I think calling NeXTstep with 16MB "quite nice" is still being generous though - my recollections of 16MB NeXTstep was "swaps less". With 32MB it started to get pretty reasonable but, gosh, that was $1000+ of RAM back then.
I knew people who ran NeXTstep on Intel but they tended to turn into early linux adopters. X11+fvwm (or, later, AfterStep) was more responsive on their hardware.
By the time NeXTSTEP 3.3 came out (and NeXT basically left the OS game, instead focusing on OpenStep and WebObjects) the performance gap was a lot smaller, at least if you could afford a really nice computer. When it reemerged as a product 5 years later as OS/X 10.0 that gap had evaporated. Suddenly it was living in its moment, 12 years after its debut.
And hell, now the OS that I used to curse for its unreasonable hardware demands comfortably runs on my cell phone. Technology is funny that way.
I love OS X and Unix but it has always felt dog slow to me. It was infuriating in PPC days but it's still frustrating even in current hardware.
As someone who has worked on Macs with kilobytes of RAM and was flabbergasted by BeOS, waiting for my computer to respond seems unacceptable.
If they finish their Wifi implementation (it works, even with WPA2 now, just doesn't have a complete GUI), bring their Webkit browser up-to-date, and get a bit more stability it would be a pleasant system to use on a day-to-day basis.
I honestly think if it had the momentum of a larger community, it could easily be open source's triumph on the desktop.
"... now I'm a vulture capi... er, venture capitalist."
His story of getting himself fired from Apple by Scully a bit prior to that was also quite amusing and revealing of his frank nature.
I also loved Bill Atkinson's story at 18:17 paraphrasing Steve Jobs convincing him to drop out of med school and join Apple:
"You know how fun it is to surf on the front edge of a wave? Now think about the poor guy dog-paddling on the tail end of the same wave... it's no fun! You go the same distance, you know, but this guy's just eating the back of the wave; this guy's surfing. Come to Apple and surf!"
And finally, for those wanting to skip the bulk of the video and go straight to the part with Gassée discussing NeXT and Be, it's at 58:22.
Company worth $80 million.
Buyout offer at $125 million.
Hold out for $275 million.
Get nothing and go out of business.
I personally don't like going to sites without knowing their address. To expand shortlinks, try http://longurl.org/
Was BeOS worse than NeXTSTEP?
It was amazing technology. It is still the most reliable desktop I ever had. It could play more videos/music etc at once with my 32 megs on a dual 66 ppc than my macbook pro w/ 4 gigs on memory and some 2.8ish dual core processor does today. Anecdote-
I used to blow people ( including myself away ) by downloading large video and music files from the internet and then launching them and playing on a loop. I got up to play in the 20-30 range at once without any lack of responsiveness and with no stutter etc.
It felt better when you were using the GUI, and maybe this is all that matters to most users, but the implementation was significantly poorer compared to NeXTSTEP.
A real time OS has some advantages as a desktop OS, everything is instant and without lag, but the total throughput of the system is significantly lower. To make a stupid car analogy, a RTOS is a Ferrari while a conventional OS is a massive off-road vehicle. Personally I think a RTOS is the way to go for a desktop OS.
I sometimes don't understand people's hate for the playbook, it is a wonderful tablet. IMHO it is a much better tablet than kindle fire (which is inspired heavily by the playbook btw), with TAT integration and two extra cameras and an active bezel. I time and time again feel that Google should buy rim and integrate QNX and Android.
But there seem to be some other interesting examples, like QNX: http://en.wikipedia.org/wiki/List_of_real-time_operating_sys...
RTLinux is more widely used than RTAI or Xenomai, but it's a completely different model, it's more like a hypervisor. Again, it doesn't help. While the OS provides the serivices, applications must be programmed specifically to use it.
And how "fast" and "real time" a desktop feels usually has little to do with the technically real time (or lack of thereof) of the kernel.
That said, Plan 9 has an interesting implementation of real time capabilities: http://doc.cat-v.org/plan_9/real_time/
QNX is a still very popular realtime OS in the embedded world today. Btw, the OS run by the BlackBerry devices is QNX. Usually you'll find QNX in control systems like avionics.
One of the major deficiencies in Apple's OS at the time was lack of real multi user capabilities; like Windows pre XP. NeXT, coming from BSD, had this built in but BeOS was pretty much a single user OS
Incidentally, up until version 7.6, the Mac OS was simply called "System", as in "System 6" and "System 7". (Systems 1-5 were rapidly iterated through and forgotten--even System 6 still ran on a Mac Plus). Mac OS 7.6 was the first version that was designed to run on clones, hence the change from generic "System" to "Mac OS".
BeOS was arguably less ready than NeXTSTEP, lacking a network stack and printing (as has been pointed out elsewhere), and had very few apps and very few users, whereas NeXT was already deployed and in use by a lot of people.
> and could have replaced MacOS the next day
Not at all. One of the main points that MacOS had to support Mac application. There is no way that Apple ever would have released a Mac operating system (be it Copland, Rhapsody or MacOS X) without a "yellow box" MacOS System 8 layer.
Didn't happen with NeXTSTEP either -- a lot of stuff, like QuickTime and Finder, had to be ported over first.
http://lowendmac.com/musings/boxes.shtml
Originally, Carbon was not in the plans at all (the yellow box corresponded to what is nowadays referred to as Cocoa). Adding that was a substantial additional task, but probably well worth it.
BeOS had a networking stack at the time; the problem was that it was notoriously buggy and incompatible with the standard POSIX APIs.
Be was working on a replacement to it when they were sold to Palm, known as BONE. You can find a bit of information about it here: http://wiki.bebits.com/page/WhatIsBone
At the time period we're talking about, it was clear that the writing was on the wall about the internet being the future, and the poor networking stack on Mac OS was going to be a major problem going forward. NeXT had a mature stack based on BSD; BeOS had an immature stack that was in dire need of being ripped up and rewritten.
On that front, NeXT's networking was a clear step forward, and Be's barely a step sideways.
I am not saying they made the wrong decision, quite the contrary I am saying they made the right one, but I was just adding some historical context to the discussion that many of us at the time thought Be was the better choice. History has proven that we where wrong, but none the less a lot of us thought Be fit the bill. I no longer hold that position given what the future became, but I would be lying if I said, I and many people saw the logic back then.
I think the best answer is BeOS wasn't as complete as NeXT. Specifically it had a fairly poor network stack (it was being retooled around the time BeOS became part of palm, no idea if it was finished)
It lacked the kind of print support MacOS and NeXT had. It lacked the same caliber of internationalization. All fixable stuff but also all stuff requiring some investment.
It also remains to be seen how well BeOS would actually scale, it was an RTOS with a UI, it could absolutely do sexy stuff on low powered machines of the day, that doesn't mean it could run a 12core dual i7 machine well. Maybe it could, we really don't know.
Certainly not Xnu.
There was indeed a PPC version of NT that was designed for CHRP. But, separate from that, Apple also seriously considered using a custom version of NT with a Mac-like GUI to replace the Mac OS. It was one of many options under consideration along with Be, NeXT, and Solaris.