At work I run Firefox on suse, and IE6 on WinXP on Vmware Server on suse:).
The WinXP virtual machine uses the legit license key which came with the Dell hardware. So even if you're a linux fan, if you have an OEM windows bundled with your hardware, don't throw it away - it might come in handy.
If stuff happens I just revert the whole WinXP to a known point. I save persistent stuff to a network share on suse.
Now the sad thing is, after a few days Firefox ends up using more RAM than the entire WinXP virtual machine running IE, and other stuff. If the firefox bunch can't get memory management right they should allow users to open new browser windows in a different _process_ so we can kill/close them without affecting the others. Right now, you can't do that, so if stuff happens, everything has to be closed. Even IE allows you to do use separate processes.
Just like you have to delete spam that sneaks through your antispam filters everyday.
What next, a list of 100 "updates" where I have to keep unchecking every one of them except the one I want?
Looks a bit like spam to me:).
I already _dislike_ Quicktime more than Windows Media Player in terms of _evilness_ and general "quality". Quicktime crashes more often than Windows Media Player. If something written in C/C++ crashes, it usually means that "an attacker can execute arbitrary code of the attacker's choice".
"This kind of thinking is not natural for most people. It's not natural for engineers. Good engineering involves thinking about how things can be made to work; the security mindset involves thinking about how things can be made to fail"
In my opinion, good engineering involves thinking that things _will_ eventually fail, how it can be made to fail _safely_ if possible and figuring out what the acceptable risk is given the cost. Modern engineers don't normally design stuff to last for 1000 years (some of it might last that long - distribution curves and all that).
Even if it's a convicted rapist, to me the "sex offender list" thing is _WRONG_.
If someone bashed me up he doesn't end up on a "public violent offender list" where he gets permanently ostracised from society for the rest of his life. Nor should he. Even if I don't forgive him, it's society's job to make a good enough attempt at pretending so that he gets a _chance_ to start anew or show how bad he really is (in which case you can justify putting him in for longer the next time).
Not all criminals return to their old ways after serving their time/punishment. If you all hate them so much and are unable to get criminals to turn over a new leaf, just imprison them for life or execute them. Stop pretending you're in the land of the free.
Putting people on permanent "witch hunt" lists for whatever reason sure sounds like "cruel and unusual punishment" to me.
As for the child porn. If that's so bad, perhaps we should be jailing people who watch Hollywood movies which depict people getting _killed_ or raped.
If it's because the killing is not for real, then how about if people started rendering child porn images on computers? Would that be legal?
And what if the killing is for real? Does that mean we should jail people who watched video clips of Budd Dwyer shooting himself?
Or does it depend on what they are thinking when they are watching it? Like you know - Thought Crime?
The problem is due to the WiFi "consortium" getting things wrong from the start, and continuing to get things wrong later.
Because of their bungling it's hard to: 1) Have authenticated users + secure wifi connection. 2) Have anonymous users + secure wifi connection (a bit like anonymous https - only the server needs to be validated).
Anyway years ago I proposed a TLD (.here) to ICANN and IETF.
One of the benefits was people could have a standard way to find out more about the local network they are using. e.g. http://here/ http://whats.here/ etc.
But I guess the ICANN was too busy approving "yet another.com" TLDs like.biz and.info.
I like the analogy too - but I'm biased since I thought of it:).
I don't know about the Java libraries now, but years ago (Java 1.2?) I seem to recall that doing _common_ things that would be simple in other languages or language libraries are often more complicated in Java. I wasn't using Java then either but my colleagues were and they seemed to have to do a lot more work:).
For example - finding out how many rows a DB query just returned. Years ago it was more work in JDBC - something like move to last row of result set and get row counter value.
It's trivial in other language libs (e.g. perl DBI). Sure some databases don't support it, but I can choose to not support those databases. Java made you "pay" even if you weren't going to use those DBs.
Back then apparently http cookie handling in java was really crap - my colleague was struggling when trying to _make_ http requests and deal with cookies read/write etc.
Whereas that sort of thing was easy with perl LWP and HTTP::Cookies.
IIRC they had lots of other probs too, but I can't remember the details - was not my prob - I was using perl for my stuff. Looking at them, I sure wasn't going to switch my stuff to java.
I guess it helps java programmers feel like they deserve their pay because of all the work they are doing;).
Strange it bothers you so much that you have to keep bringing it up.
I assume that people here would have realized it from the start or learnt to not click on the link very quickly, or turned sigs off[1], or figured some geeky solution - e.g. got their browser to modify how my sig/posts appears to them.
This is Slashdot - News for nerds. I expect people here to not be so incompetent, otherwise maybe they really should click on my sig, logout and post as AC aka "beneath your current threshold".
[1] Go to: http://slashdot.org/my/comments (log in first if you have to) then select "Disable Sigs" and click on Save.
"My daughter is 26 and I noticed that she and her friends value friendships more than careers"
She might be a typical female, or just sane:). Most girls are more interested in relationships than things. They tend to be more sane about that sort of thing than guys.
So no surprise if fewer girls go into IT or the sciences. Or fewer girls end up as CEOs. It can be lonely at the top. When you are a boss AND start laying off people, don't be surprised if people aren't quite as friendly, or if they are friendly it seems a bit fake.
Anyway, a relative who has young daughters said this - from his observation of his friends etc, his daughters' lifestyles are likely to be more dependent on who they marry than what job or degree they get. Heh I guess he assumes they'll get married.
"Do you realize the massive amount of work required to run a company? " Of course, that's why I'm not running a company.
"Do you understand the job security you have as an employee of a company?" Yes. It's in the contract.
"I have to be multi-talented, multi-disciplined, multi-tasking, and multi-personality"
Multipersonality? You should avoid that term - I read that as either two-faced or crazy;).
"don't tell me that I don't deserve it"
If you are actually an owner you take more risk, so yes you should get more reward.
But if you are one of those slash and burn CEOs, sorry, I don't really see the "value add" to the company.
They are the reason why there is no job loyalty anymore, why there is disrespect and cynicism, and why people demand more than they worth (those CEOs set the example - they damage a company and earn millions for it).
Basically I don't have loyalty to companies. I have loyalty to _people_ (bosses, colleagues, friends etc) who have earned my loyalty.
If there's a new boss, I don't see why I should automatically be loyal to the new one, just because it's allegedly the "same company".
And, sure sales is hard. But "trump you anyday"? Hey, if you were such a good sales person where you really are 100% "responsible for procuring 100% of the business" then why don't you sack everyone else and keep 100% of the money from sales?
Nowadays with so many bosses like you who "trump" employees any day, it makes little sense for employees to be willing to do a massive amount of work for the company for just the usual paycheck.
After all it is clear that it's YOUR company, not theirs. Maybe in the good old days it was their company too, but times have definitely changed.
A fair day's work for a fair day's wage. You take the extra risk, so you get the "trump" money. There's no good reason why employees should be _expected_ to put in _extra_ hours and work _every_ day. A few times a year sure - stuff happens and all that. But not all the time. Otherwise it just means management is _lying_ about the working hours as per the job _contract_. If 72 hour weeks are required to compete against China etc, put those working hours in the contract and job description. If those working hours are illegal, go figure.
Perhaps if people were working for a cooperative and not a company then things could be different. Cooperatives don't make bosses a lot of money (and so they don't tend to spring into existence as often), but many cooperatives actually do quite well. Makes me wonder how much money "professional" CEOs should really be paid when it isn't their money invested in the company.
The dangerous hackers aren't those that cause things to go down. So how are you going to notice those? It's hard for a hacker to bring the whole of google.com down, especially since it's "sharded" - you could just have different teams in charge of keeping their respective shards up and give bonuses based on uptime (factoring out externalities beyond their control) - then regularly adopt the best practices and ideas from the top groups.
Good luck with the IDS/IPS when your employees also use encryption (ssh, https etc), leave jobs running overnight, remotely come in to do work after hours etc.
I personally think that Google really aren't doing anything that wonderful with respect to security. They just haven't been unlucky so far (or maybe they have been unlucky, and just nobody knows about it yet;) ).
You see lots of people who obviously aren't good drivers, but they manage to avoid crashes _most_ of the time.
I've seen companies with pretty crap security (I used to be in the IT security line), but in most cases no hackers brought anything down there either.
Nah the first thing I thought of when I saw the countries participating was ECHELON.
I think they're probably going to scare the people to help justify spying and dubious stuff (like the spread of nasty policies to other countries).
I'm personally not afraid of all that malware, while there are lots of bots, there aren't that many hackers out there actually controlling those bots.
The fact that the cops have done little is mainly because it is not a priority. After all they could always follow the _money_. I am sure that some of the money trails will lead to their jurisdiction. Then you could also do stings.
I'm more afraid of the policies and laws that will result from this "cooperation", because they will probably try to infect other countries with the resulting mal-laws.
1) The human process could require review before commits, but is there anything preventing a person from making extra commits after a review? It might _eventually_ get spotted, but by that time it might have been exploited already. 2) Automated scripts that check security of code. I think those only work for _nonmalicious_ coders - e.g. they detect common "oops" conditions. I doubt you could have automated scripts that can detect malicious backdooring of code. Unless Google has Skynet online already;). 3) Of the few who know the search ranking stuff, what are their controls? Are they allowed to discuss it amongst themselves on Gmail?
Google's track record on security isn't very good. It's not something they are good at.
Thing is though most companies aren't good at security, most still don't get pwn3d regularly, even the big juicy targets.
I don't think he's really a geek. I think he understands geeks - he got Woz and others to make great stuff for him. Nor do I think Woz was an asshole - he was a prankster no doubt, but asshole? Nope. And Woz was one of the Top Geeks.
Thing is a lot of people need someone to yell at them to move from good to great - and many actually know that. Like those coaches hired to yell at the basketball players.
I guess those geeks stay not because they like the abuse but because they know after all that yelling, they actually produce some of the best work they have ever done (and may ever do in their lives?). AND Steve Jobs confirms it by saying it's great:).
You succeed and win, so all that pain and abuse becomes worth it.
In contrast it will be very demotivating if you get yelled at from bosses who can't tell the difference. Imagine if you start shooting 3 pointers in a row after having a bad round and unexpectedly the coach _starts_ yelling obscenities at you, when previously he was yelling encouragements...
And if a boss gives lavish praise to you when you do what you know is just an adequate job, I think it doesn't increase your confidence in that boss's ability to tell the difference. Less likely for the team to produce _great_ stuff, though it might still produce good stuff.
So I think the Wired piece misses the point by focusing on the "doing everything wrong" and not knowing what Apple are doing right. The worry is lots _more_ companies might start emulating the "doing wrong" bits without the important "doing right" bits.
I don't think we need more companies run by abusive people producing yet more crap.
Say what you like about Steve Jobs. But he has _taste_.
If CxOs are thinking of being the "the red-faced, tyrannical boss" they better not forget that important point. They're not going to do much good if they do the tyrannical part without the taste part. In fact to emulate Apple I bet the tyrannical part is optional, the taste part isn't[1]. And the taste part is _hard_ to emulate.
Jobs knows the difference between good and great. Whereas most CxOs (or people in general) can't even seem to tell the difference between good and bad:).
The typical committee might take weeks to tell you whether a piece of chocolate tastes good or not, much less even get around to the way it _looks_.
The Techs? Many of the good ones might come with great _technical_ architectures and designs - but when the customer looks at it and tries to use it, it IS a piece of crap from their PoV.
So even if the Techs at Apple don't like his abusive micromanagement, I bet they _respect_ it because Steve Jobs has taste.
They can be confident that even if he's deciding on the "curve of a monitor's corners": 1) The decision is based on making an "insanely great"[2] product (not a crony richer, or more powerful) 2) He is 90% likely to be right about what the market will like. 3) If he yells at you, it's not _just_ because he's an asshole, deep down you know know he is right - that what you just showed him is only suitable as "blah stuff" from Dell...
Many (not all) techs can accept assholes who are right most of the time.
Thing is I wonder whether it's a bit like abused spouse syndrome for them;).
[1] That said, I think a lot of people with taste AND an obsessive eye for detail tend to get very upset when stuff misses the mark. [2] Yes I know their products aren't really insanely great.
"So what mistake could one employee make to ruin it for everyone"
Get pwn3d and: a) Commit GMail/etc code secretly backdoored by a hacker. b) Leak out the search ranking and antisearch spam methods/algorithm google uses. Google's search results are already not as good as they were years ago.
If you allow some employees access through those walls to other networks, and a hacker manages to get their credentials it can start to get quite nasty.
Even if the isolation between networks is good there's also the possibility of _work_ being secretly tampered with. I'm sure there are hacker who would want to tamper with GMail or Google Desktop.
Slashdot Mirror
At work I run Firefox on suse, and IE6 on WinXP on Vmware Server on suse :).
The WinXP virtual machine uses the legit license key which came with the Dell hardware. So even if you're a linux fan, if you have an OEM windows bundled with your hardware, don't throw it away - it might come in handy.
If stuff happens I just revert the whole WinXP to a known point. I save persistent stuff to a network share on suse.
Now the sad thing is, after a few days Firefox ends up using more RAM than the entire WinXP virtual machine running IE, and other stuff. If the firefox bunch can't get memory management right they should allow users to open new browser windows in a different _process_ so we can kill/close them without affecting the others. Right now, you can't do that, so if stuff happens, everything has to be closed. Even IE allows you to do use separate processes.
Just like you have to delete spam that sneaks through your antispam filters everyday.
:).
What next, a list of 100 "updates" where I have to keep unchecking every one of them except the one I want?
Looks a bit like spam to me
I already _dislike_ Quicktime more than Windows Media Player in terms of _evilness_ and general "quality". Quicktime crashes more often than Windows Media Player. If something written in C/C++ crashes, it usually means that "an attacker can execute arbitrary code of the attacker's choice".
It's a bit like spamming people with a browser install :).
Fortunately you can opt out before you actually start downloading the browser.
Please press 1 for English. ...
Para Espanol, oprima numero dos.
Your call is important to us...
That also depends on which international region the user has set things to.
:).
For example in Arabic you read from right to left. So the order of things would make more sense the other way round.
See this:
http://www.i18nguy.com/MiddleEastUI.html
It was a pain, but seems they have a lot of money there and we wanted some of it
A Gnu herd for burgers.
Make sure you get some of that open sauce stuff.
It's in Beta we'll fix that in the release ;).
"So is the glass half-full or half-empty??"
;).
It's overengineered - someone has more glass than needed
suddenoutbreakofcommonsense :).
Dunno what that District Court was thinking though...
"This kind of thinking is not natural for most people. It's not natural for engineers. Good engineering involves thinking about how things can be made to work; the security mindset involves thinking about how things can be made to fail"
In my opinion, good engineering involves thinking that things _will_ eventually fail, how it can be made to fail _safely_ if possible and figuring out what the acceptable risk is given the cost. Modern engineers don't normally design stuff to last for 1000 years (some of it might last that long - distribution curves and all that).
"In Taiwan I've seen inkjet printers in shops running off four 500ml bottles of ink, which must be cheap as hell"
;).
In the stifling lawless environment of Taiwan it's no surprise that Taiwan is nowhere in technology
I'm sure the US would soon convince them to "harmonize their laws" with "the rest of the civilized countries" (e.g. UKUSA).
"Respect My Monopoly!"
"will be labeled as sex offenders for life"
Even if it's a convicted rapist, to me the "sex offender list" thing is _WRONG_.
If someone bashed me up he doesn't end up on a "public violent offender list" where he gets permanently ostracised from society for the rest of his life. Nor should he. Even if I don't forgive him, it's society's job to make a good enough attempt at pretending so that he gets a _chance_ to start anew or show how bad he really is (in which case you can justify putting him in for longer the next time).
Not all criminals return to their old ways after serving their time/punishment. If you all hate them so much and are unable to get criminals to turn over a new leaf, just imprison them for life or execute them. Stop pretending you're in the land of the free.
Putting people on permanent "witch hunt" lists for whatever reason sure sounds like "cruel and unusual punishment" to me.
As for the child porn. If that's so bad, perhaps we should be jailing people who watch Hollywood movies which depict people getting _killed_ or raped.
If it's because the killing is not for real, then how about if people started rendering child porn images on computers? Would that be legal?
And what if the killing is for real? Does that mean we should jail people who watched video clips of Budd Dwyer shooting himself?
Or does it depend on what they are thinking when they are watching it? Like you know - Thought Crime?
The problem is due to the WiFi "consortium" getting things wrong from the start, and continuing to get things wrong later.
.com" TLDs like .biz and .info.
Because of their bungling it's hard to:
1) Have authenticated users + secure wifi connection.
2) Have anonymous users + secure wifi connection (a bit like anonymous https - only the server needs to be validated).
Anyway years ago I proposed a TLD (.here) to ICANN and IETF.
One of the benefits was people could have a standard way to find out more about the local network they are using. e.g. http://here/ http://whats.here/ etc.
But I guess the ICANN was too busy approving "yet another
I like the analogy too - but I'm biased since I thought of it :).
:).
;).
I don't know about the Java libraries now, but years ago (Java 1.2?) I seem to recall that doing _common_ things that would be simple in other languages or language libraries are often more complicated in Java. I wasn't using Java then either but my colleagues were and they seemed to have to do a lot more work
For example - finding out how many rows a DB query just returned. Years ago it was more work in JDBC - something like move to last row of result set and get row counter value.
It's trivial in other language libs (e.g. perl DBI). Sure some databases don't support it, but I can choose to not support those databases. Java made you "pay" even if you weren't going to use those DBs.
Back then apparently http cookie handling in java was really crap - my colleague was struggling when trying to _make_ http requests and deal with cookies read/write etc.
Whereas that sort of thing was easy with perl LWP and HTTP::Cookies.
IIRC they had lots of other probs too, but I can't remember the details - was not my prob - I was using perl for my stuff. Looking at them, I sure wasn't going to switch my stuff to java.
I guess it helps java programmers feel like they deserve their pay because of all the work they are doing
I must say Java sure runs a lot faster now.
Strange it bothers you so much that you have to keep bringing it up.
I assume that people here would have realized it from the start or learnt to not click on the link very quickly, or turned sigs off[1], or figured some geeky solution - e.g. got their browser to modify how my sig/posts appears to them.
This is Slashdot - News for nerds. I expect people here to not be so incompetent, otherwise maybe they really should click on my sig, logout and post as AC aka "beneath your current threshold".
[1] Go to: http://slashdot.org/my/comments (log in first if you have to) then select "Disable Sigs" and click on Save.
"My daughter is 26 and I noticed that she and her friends value friendships more than careers"
:). Most girls are more interested in relationships than things. They tend to be more sane about that sort of thing than guys.
She might be a typical female, or just sane
So no surprise if fewer girls go into IT or the sciences. Or fewer girls end up as CEOs. It can be lonely at the top. When you are a boss AND start laying off people, don't be surprised if people aren't quite as friendly, or if they are friendly it seems a bit fake.
Anyway, a relative who has young daughters said this - from his observation of his friends etc, his daughters' lifestyles are likely to be more dependent on who they marry than what job or degree they get. Heh I guess he assumes they'll get married.
"Do you realize the massive amount of work required to run a company? "
;).
Of course, that's why I'm not running a company.
"Do you understand the job security you have as an employee of a company?"
Yes. It's in the contract.
"I have to be multi-talented, multi-disciplined, multi-tasking, and multi-personality"
Multipersonality? You should avoid that term - I read that as either two-faced or crazy
"don't tell me that I don't deserve it"
If you are actually an owner you take more risk, so yes you should get more reward.
But if you are one of those slash and burn CEOs, sorry, I don't really see the "value add" to the company.
They are the reason why there is no job loyalty anymore, why there is disrespect and cynicism, and why people demand more than they worth (those CEOs set the example - they damage a company and earn millions for it).
Basically I don't have loyalty to companies. I have loyalty to _people_ (bosses, colleagues, friends etc) who have earned my loyalty.
If there's a new boss, I don't see why I should automatically be loyal to the new one, just because it's allegedly the "same company".
And, sure sales is hard. But "trump you anyday"? Hey, if you were such a good sales person where you really are 100% "responsible for procuring 100% of the business" then why don't you sack everyone else and keep 100% of the money from sales?
Nowadays with so many bosses like you who "trump" employees any day, it makes little sense for employees to be willing to do a massive amount of work for the company for just the usual paycheck.
After all it is clear that it's YOUR company, not theirs. Maybe in the good old days it was their company too, but times have definitely changed.
A fair day's work for a fair day's wage. You take the extra risk, so you get the "trump" money. There's no good reason why employees should be _expected_ to put in _extra_ hours and work _every_ day. A few times a year sure - stuff happens and all that. But not all the time. Otherwise it just means management is _lying_ about the working hours as per the job _contract_. If 72 hour weeks are required to compete against China etc, put those working hours in the contract and job description. If those working hours are illegal, go figure.
Perhaps if people were working for a cooperative and not a company then things could be different. Cooperatives don't make bosses a lot of money (and so they don't tend to spring into existence as often), but many cooperatives actually do quite well. Makes me wonder how much money "professional" CEOs should really be paid when it isn't their money invested in the company.
The dangerous hackers aren't those that cause things to go down. So how are you going to notice those? It's hard for a hacker to bring the whole of google.com down, especially since it's "sharded" - you could just have different teams in charge of keeping their respective shards up and give bonuses based on uptime (factoring out externalities beyond their control) - then regularly adopt the best practices and ideas from the top groups.
;) ).
Good luck with the IDS/IPS when your employees also use encryption (ssh, https etc), leave jobs running overnight, remotely come in to do work after hours etc.
I personally think that Google really aren't doing anything that wonderful with respect to security. They just haven't been unlucky so far (or maybe they have been unlucky, and just nobody knows about it yet
You see lots of people who obviously aren't good drivers, but they manage to avoid crashes _most_ of the time.
I've seen companies with pretty crap security (I used to be in the IT security line), but in most cases no hackers brought anything down there either.
The insects are calling.
Nah the first thing I thought of when I saw the countries participating was ECHELON.
I think they're probably going to scare the people to help justify spying and dubious stuff (like the spread of nasty policies to other countries).
I'm personally not afraid of all that malware, while there are lots of bots, there aren't that many hackers out there actually controlling those bots.
The fact that the cops have done little is mainly because it is not a priority. After all they could always follow the _money_. I am sure that some of the money trails will lead to their jurisdiction. Then you could also do stings.
I'm more afraid of the policies and laws that will result from this "cooperation", because they will probably try to infect other countries with the resulting mal-laws.
1) The human process could require review before commits, but is there anything preventing a person from making extra commits after a review? It might _eventually_ get spotted, but by that time it might have been exploited already. ;).
2) Automated scripts that check security of code. I think those only work for _nonmalicious_ coders - e.g. they detect common "oops" conditions. I doubt you could have automated scripts that can detect malicious backdooring of code. Unless Google has Skynet online already
3) Of the few who know the search ranking stuff, what are their controls? Are they allowed to discuss it amongst themselves on Gmail?
Google's track record on security isn't very good. It's not something they are good at.
Thing is though most companies aren't good at security, most still don't get pwn3d regularly, even the big juicy targets.
I don't think he's really a geek. I think he understands geeks - he got Woz and others to make great stuff for him.
:).
Nor do I think Woz was an asshole - he was a prankster no doubt, but asshole? Nope. And Woz was one of the Top Geeks.
Thing is a lot of people need someone to yell at them to move from good to great - and many actually know that. Like those coaches hired to yell at the basketball players.
I guess those geeks stay not because they like the abuse but because they know after all that yelling, they actually produce some of the best work they have ever done (and may ever do in their lives?). AND Steve Jobs confirms it by saying it's great
You succeed and win, so all that pain and abuse becomes worth it.
In contrast it will be very demotivating if you get yelled at from bosses who can't tell the difference. Imagine if you start shooting 3 pointers in a row after having a bad round and unexpectedly the coach _starts_ yelling obscenities at you, when previously he was yelling encouragements...
And if a boss gives lavish praise to you when you do what you know is just an adequate job, I think it doesn't increase your confidence in that boss's ability to tell the difference. Less likely for the team to produce _great_ stuff, though it might still produce good stuff.
So I think the Wired piece misses the point by focusing on the "doing everything wrong" and not knowing what Apple are doing right. The worry is lots _more_ companies might start emulating the "doing wrong" bits without the important "doing right" bits.
I don't think we need more companies run by abusive people producing yet more crap.
Say what you like about Steve Jobs. But he has _taste_.
:).
;).
If CxOs are thinking of being the "the red-faced, tyrannical boss" they better not forget that important point. They're not going to do much good if they do the tyrannical part without the taste part. In fact to emulate Apple I bet the tyrannical part is optional, the taste part isn't[1]. And the taste part is _hard_ to emulate.
Jobs knows the difference between good and great. Whereas most CxOs (or people in general) can't even seem to tell the difference between good and bad
The typical committee might take weeks to tell you whether a piece of chocolate tastes good or not, much less even get around to the way it _looks_.
The Techs? Many of the good ones might come with great _technical_ architectures and designs - but when the customer looks at it and tries to use it, it IS a piece of crap from their PoV.
So even if the Techs at Apple don't like his abusive micromanagement, I bet they _respect_ it because Steve Jobs has taste.
They can be confident that even if he's deciding on the "curve of a monitor's corners":
1) The decision is based on making an "insanely great"[2] product (not a crony richer, or more powerful)
2) He is 90% likely to be right about what the market will like.
3) If he yells at you, it's not _just_ because he's an asshole, deep down you know know he is right - that what you just showed him is only suitable as "blah stuff" from Dell...
Many (not all) techs can accept assholes who are right most of the time.
Thing is I wonder whether it's a bit like abused spouse syndrome for them
[1] That said, I think a lot of people with taste AND an obsessive eye for detail tend to get very upset when stuff misses the mark.
[2] Yes I know their products aren't really insanely great.
"So what mistake could one employee make to ruin it for everyone"
Get pwn3d and:
a) Commit GMail/etc code secretly backdoored by a hacker.
b) Leak out the search ranking and antisearch spam methods/algorithm google uses. Google's search results are already not as good as they were years ago.
That's fine if the walls are 100%.
If you allow some employees access through those walls to other networks, and a hacker manages to get their credentials it can start to get quite nasty.
Even if the isolation between networks is good there's also the possibility of _work_ being secretly tampered with. I'm sure there are hacker who would want to tamper with GMail or Google Desktop.
Or confidential information leaking out.