6 ms·

Diary of a Disaster: General Magic Goes Poof (2004)

47 points by protomyth 11y ago

thoughtsimple 11y ago
No mention of the disastrous "cleaning up" bug when you pulled the PCMCIA RAM card. That was the final straw for me. It could take 30-45 minutes before you could use your device again after pulling the card.
- blacksqr 11y ago
  Ah, yes. Also, if you yanked the Envoy battery the wrong way before proper shutdown, total data loss. (The Envoy battery was not internal, but an integrated part of the case.)
keithwinstein 11y ago
Some more positive memories: - Magic Cap came from the early-1990s expectation that the future involved intelligent agents that would travel around the network, do work (e.g. searching for the best price on a flight or new shoes), and then report back when the user was connected again. In practice, this meant that you could send email to another Magic Cap user that included executable code, like an animation or a little game, that would run in a sandbox on the user's device. Even today, we still don't really have this -- you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it. - The Motorola Envoy ran on Motorola's low-bitrate, high-power (and now defunct) Ardis wireless network. It was robust enough to work from within an airplane!! Because almost nobody had them (in 1995), the flight attendants would not stop you from sending and receiving email wirelessly while flying. Of course today, even if Ardis still existed and still worked from 30,000 flight, they would be telling you to put your phone into airplane mode... - I did not have a Motorola Envoy, but I did have a Sony Magic Link that I used for email from 1995-1996. It was amazing! You could leave it by your bed, and when you woke up in the morning, it would make a (free, local) call to AT&T PersonaLink to exchange email. I have fond memories of lying in bed in my PJs in early 1996, having just woken up, and flirting via email with a high-school friend who had already arrived at school at the computer lab. This all seemed like an incredible luxury at the time. Of course "checking email when you first wake up" became much more common with the end of per-minute fees from major ISPs (AOL ended theirs in December 1996), the rise of laptops and pervasive Wi-Fi (~2001), and eventually the rise of cellular handhelds with email (~2005).
- arethuza 11y ago
  "you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it." You say that like it's a bad thing? NB I can see that it might make sense to allow "executable" messages - but they potential for security and/or privacy problems look pretty bad.
  - protomyth 11y ago
    You could stick Postscript code in a NeXTMail and do some pretty funky things. Not sure how sandboxes it was. I'm not sure running agents is all that more a problem than the JavaScript in a web browser. The agent folks were definitely thinking about security and even CPU usage.
- wmf 11y ago
  We don't have mobile agents today because we have decent Internet access (even on phones) so it's fast enough and much easier to use RPC-like communication. In retrospect, the whole mobile agent concept was a bet against the Internet.
- Tloewald 11y ago
  > Even today, we still don't really have this -- you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it. We did have it -- it was called VBScript in Outlook and it was one of the biggest (actually probably the biggest) security disasters in the history of the internet.
  - blacksqr 11y ago
    And Tcl has had a safe interpreter feature and a sophisticated policy mechanism from early on. It was used in the Tcl browser plugin, which routinely downloaded code from the Web and executed it. The plugin is gone but the safe interpreter and policy interface are still perfectly functional. Tcl also had several agent extensions back in the 1990's.
protomyth 11y ago
"Steve Wozniak, the cofounder of Apple, has signed away several of his patents on infrared control technologies to General Magic" "We find that a phone link is actually faster than IR! The poor IR performance is particularly wounding to Woz, who signed away his IR patents and technology to General Magic." a lot of bug discussion and ignoring Woz "Wozniak is visibly upset that the product has not been fixed and that the stock market is so gullible. Perhaps be is also miffed that be was not even offered stock for his gift of the IR patents some three years ago." If they treated Woz they way this article says, then I cannot help but think they were probably going to fail. Its a true shame because I still think agents would be a very interesting way to organize programs even though the remote communication part has been taken over by http-based APIs.
stretchwithme 11y ago
How amazing is it that nearly every problem encountered with those early devices has been solved in the intervening years?
- smacktoward 11y ago
  Given twenty years and the combined engineering budgets of every major technology company during that time, General Magic could probably have solved them all too :-D
  - stretchwithme 11y ago
    I don't know about that. Many competitors trying out often conflicting ideas probably makes things go a lot faster than one company could ever go.
mwcampbell 11y ago
I'm surprised that such a buggy piece of software was developed by Bill Atkinson and Andy Hertzfeld, who made QuickDraw and the original Macintosh Toolbox work in even less RAM (128K, soon upgraded to 512K). I wonder what caused the later project to go so badly.
- Aloha 11y ago
  See the "Second System Effect" in the mythical man month. https://en.wikipedia.org/wiki/The_Mythical_Man-Month#The_second-system_effect https://en.wikipedia.org/wiki/The_Mythical_Man-Month#The_sec...
  - Esau 11y ago
    That explains IPv6 and uEFI/EFI.
    - Aloha 11y ago
      I'd be curious to hear your thoughts on both of those things - I've considered IPv6 to be elegant even, as elegant as v4 - EFI I don't know enough about to have a strong opinion.
- smacktoward 11y ago
  The article hints at one explanation, namely GM's consortium of BigCo backers being so huge that it could never really come to consensus on a single vision of what exactly GM was supposed to be developing. So they end up getting pulled in a dozen different directions chasing the particular needs of each backer, instead of putting all their energy behind one product.
  - blacksqr 11y ago
    I think that's pure ass-covering. The reality was they just couldn't deliver a product that wasn't riddled with bugs, leaving the consortium partners scrambling to integrate the OS code with their hardware, and trying to raise hell to fix the situation. All they got from GM was rainbows blown up their *. (I worked for Motorola at the time, see comments elsewhere.)
dools 11y ago
"Early Summer, 1995 - I advise a major brokerage house not to invest in General Magic. They invest anyway. I hope they don't wind up losing a large chunk of cash." Isn't that insider trading?
- lmm 11y ago
  "Insider trading" is a loosely-defined phrase, but the regulations most people think of only apply to publicly traded stocks.
blacksqr 11y ago
Wow, brought back a lot of memories. I worked for Motorola on the Envoy team. It was my first encounter with Reality Distortion Field engineering. I remember their software engineers were literally called "wizards" as their job title. They developed a custom fully object-oriented programming language to write the OS in, and confidently assured us that it was so advanced that the OS would ultimately be delivered with literally zero bugs. This in the face of build after build of buggy, unstable, backward-incompatible code. I remember the change of line terminator from semicolon to colon (or vice versa?) when the product was already months past its original ship date. Good times.
- mwcampbell 11y ago
  Can you tell us more about the programming language and operating system? For example, how did memory management work? Was it based on garbage collection, reference counting, or still manual memory management? And what about persistent storage? I gather from the OP that fragmentation of persistent storage was a big problem.
  - blacksqr 11y ago
    We weren't allowed access to the OS code itself, just the developer API, so I can't give detailed info about the implementation. I was the chief QA engineer for the Envoy team, but I was locked out of bug discussion meetings with the GM people. Everything was handled at a very high level. I remember they were very proud of the fact that the programming language was fully object oriented, and drank deep of the OO kool-aid. So no C-style direct memory access AFAIR.
pavlov 11y ago
Here's a Feb 1994 article from Byte that offers some additional insight to the General Magic product, and has some screenshots too: http://www.guidebookgallery.org/articles/justlikemagic http://www.guidebookgallery.org/articles/justlikemagic
smacktoward 11y ago
Oct. 7, 1994 - By this time, the magic is wearing thin, and Woz has gone from exhilaration to borderline disgust. Ten days ago, he was euphorically calling the Magic Link the best computer consumer product ever made. Today, be fears his friends have made horrible engineering mistakes. I watch as Woz goes out to buy still more Magic Links to try to discover the source of the problems so he can help his friends fix them. One more story to add to the "Woz: a real mensch" pile.
- protomyth 11y ago
  Woz, Jay Miner, and Seymour Cray were my first computer heroes. It still is amazing reading stories like this to get a feel for their abilities and determination. // ok Apple WTF are we doing trying to autocorrect Woz to Wiz?
mrpippy 11y ago
Here's a developer introduction to Magic Cap from 1995: http://www.mactech.com/articles/mactech/Vol.11/11.05/MakingMagic/index.html http://www.mactech.com/articles/mactech/Vol.11/11.05/MakingM...
- mwcampbell 11y ago
  So the persistent storage was RAM? How did anyone think that was a good idea? Was flash just too expensive?
  - bigtones 11y ago
    NAND flash was not commercially available at that time.
  - thescriptkiddie 11y ago
    Yes.
  - Gibbon1 11y ago
    My memory (heh) fails me somewhat. But I think the only Flash available in 1995 was NOR Flash. So very slow, expensive, unavailable in reasonable sizes, only limited numbers of writes. When I said very slow I mean glacially slow. I have worked on systems with battery backed RAM. Works fairly well, however a naive reliance on it will result in data corruption and tears. (Hint reset or a program crash can come at any moment)
  - Tloewald 11y ago
    I repeatedly lost data with early Windows CE and Palm devices for similar reasons. The Newton, on the other hand, was persistent even with dead batteries -- I never lost data on a Newton, even the dreadful (and expensive) original Messagepad.
  - protomyth 11y ago
    Flash was later and bubble memory had faded. Magic Cap, Newton, and Palm Did the battery backed RAM. The Newton had two batteries to keep RAM alive (one main battery and a watch battery to handle longer term / main battery switch). The OS was on an EPROM.
Esau 11y ago
I remember they also offered a Magic Cap application for Windows. I never used it though, so I can't comment on how well it worked.
- protomyth 11y ago
  It was about par for the course for applications at that time (a couple of crashes when connecting), but it didn't really do a whole lot either. I used it for two months then they announced the shutdown.
joezydeco 11y ago
Anyone else remember the pivot afterward into a voice recognition product?
fernly 11y ago
A good story. Pity it is marred by a number of uncorrected OCR errors: many "h" scanned as "b", 1/0/5 scanned as I/O/S.

47 points by protomyth 11y ago

No mention of the disastrous "cleaning up" bug when you pulled the PCMCIA RAM card. That was the final straw for me. It could take 30-45 minutes before you could use your device again after pulling the card.

blacksqr 11y ago
Ah, yes. Also, if you yanked the Envoy battery the wrong way before proper shutdown, total data loss. (The Envoy battery was not internal, but an integrated part of the case.)

keithwinstein 11y ago

Some more positive memories: - Magic Cap came from the early-1990s expectation that the future involved intelligent agents that would travel around the network, do work (e.g. searching for the best price on a flight or new shoes), and then report back when the user was connected again. In practice, this meant that you could send email to another Magic Cap user that included executable code, like an animation or a little game, that would run in a sandbox on the user's device. Even today, we still don't really have this -- you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it. - The Motorola Envoy ran on Motorola's low-bitrate, high-power (and now defunct) Ardis wireless network. It was robust enough to work from within an airplane!! Because almost nobody had them (in 1995), the flight attendants would not stop you from sending and receiving email wirelessly while flying. Of course today, even if Ardis still existed and still worked from 30,000 flight, they would be telling you to put your phone into airplane mode... - I did not have a Motorola Envoy, but I did have a Sony Magic Link that I used for email from 1995-1996. It was amazing! You could leave it by your bed, and when you woke up in the morning, it would make a (free, local) call to AT&T PersonaLink to exchange email. I have fond memories of lying in bed in my PJs in early 1996, having just woken up, and flirting via email with a high-school friend who had already arrived at school at the computer lab. This all seemed like an incredible luxury at the time. Of course "checking email when you first wake up" became much more common with the end of per-minute fees from major ISPs (AOL ended theirs in December 1996), the rise of laptops and pervasive Wi-Fi (~2001), and eventually the rise of cellular handhelds with email (~2005).

arethuza 11y ago
"you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it." You say that like it's a bad thing? NB I can see that it might make sense to allow "executable" messages - but they potential for security and/or privacy problems look pretty bad.
- protomyth 11y ago
  You could stick Postscript code in a NeXTMail and do some pretty funky things. Not sure how sandboxes it was. I'm not sure running agents is all that more a problem than the JavaScript in a web browser. The agent folks were definitely thinking about security and even CPU usage.
wmf 11y ago
We don't have mobile agents today because we have decent Internet access (even on phones) so it's fast enough and much easier to use RPC-like communication. In retrospect, the whole mobile agent concept was a bet against the Internet.
Tloewald 11y ago
> Even today, we still don't really have this -- you can't stick JavaScript (or a Java applet) in an email and expect the receiver to be able to run it. We did have it -- it was called VBScript in Outlook and it was one of the biggest (actually probably the biggest) security disasters in the history of the internet.
- blacksqr 11y ago
  And Tcl has had a safe interpreter feature and a sophisticated policy mechanism from early on. It was used in the Tcl browser plugin, which routinely downloaded code from the Web and executed it. The plugin is gone but the safe interpreter and policy interface are still perfectly functional. Tcl also had several agent extensions back in the 1990's.

protomyth 11y ago

"Steve Wozniak, the cofounder of Apple, has signed away several of his patents on infrared control technologies to General Magic" "We find that a phone link is actually faster than IR! The poor IR performance is particularly wounding to Woz, who signed away his IR patents and technology to General Magic." a lot of bug discussion and ignoring Woz "Wozniak is visibly upset that the product has not been fixed and that the stock market is so gullible. Perhaps be is also miffed that be was not even offered stock for his gift of the IR patents some three years ago." If they treated Woz they way this article says, then I cannot help but think they were probably going to fail. Its a true shame because I still think agents would be a very interesting way to organize programs even though the remote communication part has been taken over by http-based APIs.

stretchwithme 11y ago

How amazing is it that nearly every problem encountered with those early devices has been solved in the intervening years?

smacktoward 11y ago
Given twenty years and the combined engineering budgets of every major technology company during that time, General Magic could probably have solved them all too :-D
- stretchwithme 11y ago
  I don't know about that. Many competitors trying out often conflicting ideas probably makes things go a lot faster than one company could ever go.

mwcampbell 11y ago

I'm surprised that such a buggy piece of software was developed by Bill Atkinson and Andy Hertzfeld, who made QuickDraw and the original Macintosh Toolbox work in even less RAM (128K, soon upgraded to 512K). I wonder what caused the later project to go so badly.

Aloha 11y ago
See the "Second System Effect" in the mythical man month. https://en.wikipedia.org/wiki/The_Mythical_Man-Month#The_second-system_effect https://en.wikipedia.org/wiki/The_Mythical_Man-Month#The_sec...
- Esau 11y ago
  That explains IPv6 and uEFI/EFI.
  - Aloha 11y ago
    I'd be curious to hear your thoughts on both of those things - I've considered IPv6 to be elegant even, as elegant as v4 - EFI I don't know enough about to have a strong opinion.
smacktoward 11y ago
The article hints at one explanation, namely GM's consortium of BigCo backers being so huge that it could never really come to consensus on a single vision of what exactly GM was supposed to be developing. So they end up getting pulled in a dozen different directions chasing the particular needs of each backer, instead of putting all their energy behind one product.
- blacksqr 11y ago
  I think that's pure ass-covering. The reality was they just couldn't deliver a product that wasn't riddled with bugs, leaving the consortium partners scrambling to integrate the OS code with their hardware, and trying to raise hell to fix the situation. All they got from GM was rainbows blown up their *. (I worked for Motorola at the time, see comments elsewhere.)

dools 11y ago

"Early Summer, 1995 - I advise a major brokerage house not to invest in General Magic. They invest anyway. I hope they don't wind up losing a large chunk of cash." Isn't that insider trading?

lmm 11y ago
"Insider trading" is a loosely-defined phrase, but the regulations most people think of only apply to publicly traded stocks.

blacksqr 11y ago

Wow, brought back a lot of memories. I worked for Motorola on the Envoy team. It was my first encounter with Reality Distortion Field engineering. I remember their software engineers were literally called "wizards" as their job title. They developed a custom fully object-oriented programming language to write the OS in, and confidently assured us that it was so advanced that the OS would ultimately be delivered with literally zero bugs. This in the face of build after build of buggy, unstable, backward-incompatible code. I remember the change of line terminator from semicolon to colon (or vice versa?) when the product was already months past its original ship date. Good times.

mwcampbell 11y ago
Can you tell us more about the programming language and operating system? For example, how did memory management work? Was it based on garbage collection, reference counting, or still manual memory management? And what about persistent storage? I gather from the OP that fragmentation of persistent storage was a big problem.
- blacksqr 11y ago
  We weren't allowed access to the OS code itself, just the developer API, so I can't give detailed info about the implementation. I was the chief QA engineer for the Envoy team, but I was locked out of bug discussion meetings with the GM people. Everything was handled at a very high level. I remember they were very proud of the fact that the programming language was fully object oriented, and drank deep of the OO kool-aid. So no C-style direct memory access AFAIR.

pavlov 11y ago

Here's a Feb 1994 article from Byte that offers some additional insight to the General Magic product, and has some screenshots too: http://www.guidebookgallery.org/articles/justlikemagic http://www.guidebookgallery.org/articles/justlikemagic

smacktoward 11y ago

Oct. 7, 1994 - By this time, the magic is wearing thin, and Woz has gone from exhilaration to borderline disgust. Ten days ago, he was euphorically calling the Magic Link the best computer consumer product ever made. Today, be fears his friends have made horrible engineering mistakes. I watch as Woz goes out to buy still more Magic Links to try to discover the source of the problems so he can help his friends fix them. One more story to add to the "Woz: a real mensch" pile.

protomyth 11y ago
Woz, Jay Miner, and Seymour Cray were my first computer heroes. It still is amazing reading stories like this to get a feel for their abilities and determination. // ok Apple WTF are we doing trying to autocorrect Woz to Wiz?

mrpippy 11y ago

Here's a developer introduction to Magic Cap from 1995: http://www.mactech.com/articles/mactech/Vol.11/11.05/MakingMagic/index.html http://www.mactech.com/articles/mactech/Vol.11/11.05/MakingM...

mwcampbell 11y ago
So the persistent storage was RAM? How did anyone think that was a good idea? Was flash just too expensive?
- bigtones 11y ago
  NAND flash was not commercially available at that time.
- thescriptkiddie 11y ago
  Yes.
- Gibbon1 11y ago
  My memory (heh) fails me somewhat. But I think the only Flash available in 1995 was NOR Flash. So very slow, expensive, unavailable in reasonable sizes, only limited numbers of writes. When I said very slow I mean glacially slow. I have worked on systems with battery backed RAM. Works fairly well, however a naive reliance on it will result in data corruption and tears. (Hint reset or a program crash can come at any moment)
- Tloewald 11y ago
  I repeatedly lost data with early Windows CE and Palm devices for similar reasons. The Newton, on the other hand, was persistent even with dead batteries -- I never lost data on a Newton, even the dreadful (and expensive) original Messagepad.
- protomyth 11y ago
  Flash was later and bubble memory had faded. Magic Cap, Newton, and Palm Did the battery backed RAM. The Newton had two batteries to keep RAM alive (one main battery and a watch battery to handle longer term / main battery switch). The OS was on an EPROM.

Esau 11y ago

I remember they also offered a Magic Cap application for Windows. I never used it though, so I can't comment on how well it worked.

protomyth 11y ago
It was about par for the course for applications at that time (a couple of crashes when connecting), but it didn't really do a whole lot either. I used it for two months then they announced the shutdown.

joezydeco 11y ago

Anyone else remember the pivot afterward into a voice recognition product?

fernly 11y ago

A good story. Pity it is marred by a number of uncorrected OCR errors: many "h" scanned as "b", 1/0/5 scanned as I/O/S.