The answer is that it Just Works(TM). Apple has made Mac OS X an extremely pleasent environment to use with little things like Alt-TAB through applications, then Alt-` through an application's windows/tabs. These little things add up into a much nicer user environment. M$ strayed from it around the 98/W2k timeframe when they declared an end to MDI applications, instead claming that each document should have its own top level window. Hence the foot dragging to add tabbed browsing to IE, and the evil word behavior. Previous versions of word defaulted to MDI you can turn this back on by disabing the "windows in taskbar option". The big hoopla over tabbed browsing is yet another case of MDI's resurgence.
I think you need a better example, windows has since 3.0 (maybe longer) the standard where alt-command works on the global application space while ctrl-command works on the application space. Very few people know this, but its true. For example, Alt-f4 closes the application, ctrl-f4 closes the window. Alt-Tab cycles through applications, ctrl-tab cycles through the current applications windows.
I will agree with you though that as far as 'unix' goes OS-X is light years beyond any other 'unix' desktop. When i'm using linux one of the reasons I use KDE almost exclusivly is because they have tranditionally done a better job keeping application keystrokes consitant. While OS-X is _VERY_ nice any its pretty i'm not so sure about the keyboard usability. I can use a windows box without a mouse, i can't do that on a mac, maybe thats because i'm just not smart enough to know all the hotkeys.
If only the strong survive in the marketplace, sooner or later there will only be one.
Why would this be true? Can you see the difference between "strong" and "strongest"? In other words, that two strong competitors can thrive in the same marketplace?
Accually its simple, there are a number things that can happen.
The two companies artificially try to maintain their share of the market, neither gaining nor loosing ground. This allows both of them to thrive and make a lot of money. Coke and pepsi are examples of this. They charge $1 for sugar water, which is about $.95 profit. The customer gets screwed.
The companies go for the kill out of the idea that they can become monopolies, eventually one of the companies looses, the remaining company raises prices to anything they wan't only lowering them in markets where an upstart challenges them.
One company leverages itself and buys the other, claiming it makes them more competitive to consolidate operations. Prices are cheaper for 1 year then they start raising prices whenever they feel like it
The companies accumulate a large amount of debt trying to grab market share and another company comes along and grabs large market share by using a diffrent model or some new efficiency. This is the ideal situation, but i'm sure a study would show doesn't occor in "mature markets"
The Democrats and Republicans in this country are both the same with slight differences having nothing to do with conservatism or liberalism. The parties are more alike than different, and they are alike in the fact that they are heavily authoritarian .
Are they going to continue to have share eaten in serverspace? Yes.
Little out of touch are we? Windows is officially the largest server OS by installed base out there. It continues to grow pretty significalty by eating in the big unix servers market share. Linux is also growing at a much faster rate but they have a few years to catch up at the current growth rates, and they are also eating into the big unix servers market share. The number of people switching from windows->linux doesn't appear to be very large. All the growth is comming from unix->linux and unix->windows.
Proper application use of registry is to write local machine settings in HKLM and user settings into HKCU. Network based applications write to the servers HKLM and user specific settings to the HKCU. In fact the HKCU tree is pulled in over the network in a NT domain.
Just because you don't understand the registry rules doesn't mean that its missing all these concepts you think it should have. As another poster pointed out the registry is an extension of the old.INI files that people just like you used to bitch about because they got scattered all over the machine. Just like on linux, i have config files in/etc,/usr/local/etc, ~/, ~/.application,/var/something and a number of other places depending on which application we are talking about. Not that this is nessisary bad, except when application vendors put them in the wrong place. This is the same as the registry, when your running as administrator and you nice new application writes to HKLM when it should have written to HKLU its the same problem. You can find these applicatons in windows by running without permissions to update HKLM same as you can run linux/unix without permission to modify/etc/something.
I'm not a big VMWare user but last time I played with it, you could emulate machines with clockspeeds of CPU Mhz/(n+1) where n is the number of virtual machines. So a 1Ghz machine could have 2VMs of around 333Mhz and a 3Ghz PC could have two
Its not really that simple, I tend to use VMware as a test and development bed. I've been purchasing MP machines for years, and usually my desktop machines have two processors. Anyway, with most things I use and write there is some level of MP scalablilty, vmware isn't any diffrent, when the virtual machine is idle it doesn't take up any physical CPU. With screen savers turned off (my virtual screen's don't need saving) and such the virtual machine will consume less than 1% of the physical CPU.
What really needs to be said is common knowledge with most people who work with MP machines, asside from some of the transaction based server machines, most programs have limits to how many concurrent operations they can run at the same time. Just like its hard to extract single thread instruction parralism past a couple of operations, its hard to extract multiple thread parralism past a couple of threads. My current application is scalable to somewhere between 5 to 10 threads, past that extra processors don't do anything for me. Basically the laws of software haven't changed in the last few years but the massive scalability of web based applications have helped people forget that throwing more processors at a problem even if its scalable won't always cause the application to scale.
Usually, somewhere someone has to consolidate all those transactions going to a DB, or whatever your bottleneck is. The attempt to sell MP machines to the masses is just a cheap and lazy way to speed peoples computers up. Problem sets that are processor bound and scalable are already running on big SMP's. Personally i like SMP's, but given the choice of a CPU 2x as fast or 2 cpu's I will take the single proc any day. Having two CPU's running 2x as fast is of course the best of both worlds and I'm usually happy to spend a little more to get that.
I predict that AMD, Intel etc will make a few multiproc CPU's get out to maybe 4 cores per die and then the race to make faster cores will kick back in. Only now people will expect screaming fast 4 way chips instead of screaming fast 1 way chips.
Everyone is saying "he must not be that smart/great" if he can't get a sound card working in Linux.. My take on it is "maybe he is to busy working on useful things to waste his time getting a sound card working in linux". As someone who gets shit done at home and work, I find that you have to focus on the problem and work on it rather than being distracted by everything around you. Spending 6 hours getting proper audio support in linux is 6 hours lost that could have been spent working on your project. Today it's 6 hours on the soundcard tomorrow its three weeks figuring out why the throughput in your application is 1/10th of what it should be (my current problem) because some idiot linux kernel 'hacker' broke part of the disk subsystem in the last 10 revisions of the kernel.
A few years ago I decided to switch my desktop back to windows 2k and exceed, and I'm significantly more productive than I ever was running linux, and wasting 3 hours trying to figure out how to remap my goto-line key in the most recent version of emacs, after the developers decided to use the key for something else.
This is silly, because while I may have 50 processes with 2 to 5 threads each, they are all idle waiting around for some event. The applications are all waiting for some network or user action, the "services" are all waiting around for network or applications to access them, the OS has a few house keeping threads eating up.01% of my CPU and the rest are running in serialized application contexts. In my case all the CPU time is being eaten up running virtual machines. Maybe vmware can burn up an extra CPU doing some kind of pre processing but I fail to see how a virtual machine will benifit from more CPU threads that are slower than one fast one...
The limiting factor for any diving to any real depth (>30 feet say) is the amount of residual nitrogen in your blood stream.
Your just going to have to go back and get a cert for decompression diving, and learn to hang out decompressing
The other big drawback I see is that at depth the pressure of the water on your body is very great. That is why modern scuba uses pressure delivery systems.
I imagine that the air from this device will be generated near the pressure of the surrounding water, thereby simplifing the regulator.
It seems to me that appropriate selection of games would have a similar affect as that site. Warcraft or similar when played at full speed is going to increase your ability to make split second decisions and track lots of diffrent objects.
Back in the mid 90's my boss in response to me bitching about windows said "Its not that M$ writes good software, its that everyone else's software is worse" This was shortly after we dumped all our unix servers in favor of NT. Yes, at that time linux was around and we had even tried it for a while. I showed it running on one of my machines and my other boss managed to lock it up in under 3 minuites. The general consensus at the time where I worked? Linux was just another unix complete will all the usual unix problems.
Maybe now, but 10 years ago, a good reason to buy logitech was that they would ship a free lefty mouse with the purchase of a righty mouse. We did this at work a few times to justify the $50 mice. A righty would use the one that came in the package and a lefty would use the one that came in the mail. Two for 1...
Having a flashligh, crowbar and gloves in a car was enough to get a friend of mine taken to jail on some charge like felony posssion of buglary tools 10 years ago. Doesn't matter there was a toolbox full of other crap. The combination of the three was enough trigger some stupid law. He got out of it but thats not the point...
They can't figure out what they are doing. They stared out with a nice product, then instead of fixing it they have been playing palm works with "buzzword of month club" games.
I could give a fsck if its on a 300mhz ARM, i could give a fsck if its linux (which seems like a recipe for disaster on a handheld). Adding a non GUI thread scheduler should be easy( and not require a heavyweight OS like linux) As a developer what I wan't is a consistant API that extends rather than gets replaced every few years. Their product isn't even that good anymore. The new plams are more like crappy CE clones rather than Palm's which were always known for a unique and very stable system with loads of applications. Now days most of the applications seem like they are for palmOS 3.5 or 4, and the new ones don't scale well down to old devices. Plus palm is to busy making 100 diffrent versions of the same product to accually provide a nice range of expansion options. Then there are the tools, frankly the opensouce tools are a pain, it takes 1 day or so to get the enviroment working enough to write a hello world application. That is if your lucky or already know what to do. Writing a PalmOS 5 application with the opensource tools is a joke...
Now I just need to find a CE device with the battery life of my old Palm 505 and a similar footprint.
Cutting the fat is seen as acceptable because people assume that there is fat. That is accually a mangement failure too. It seems to indicate that there are a large number of people in the company that arn't doing their fair share of the work. This is probably true in a lot of these companies, but that accually stems from bad managment. If managment was accually keeping tabs on what everyone was doing they would know who just shows up and surfs/. all day instead of being productive. That situation would either be corrected or the person would be let go. Having people who were useful 10 years ago just sit around and surf is a bad idea for profitibility.
Hmmp. I would fly first class on a regular basis if the prices were more in line. Everytime I try to get a first class seat though it costs 5x as much as a normal seat for only about 2x the space.
Look at that table closer. Legacy mode includes "real-mode" and it does work in that mode. In "Long mode" it only supports 64-bit and 32-bit instructions. This is the mode that win64 runs in. If they wanted to run 16-bit applications they would have to either dynamically switch the processor back and forth between long and legacy, or write a software emulator to emulate the 16-bit instructions.
Accually most of the 64-bit unix's also have long=32 bits. I agree its stupid, long should have been 64-bits but now you have the "long long" basdardization which is 64-bits. At least M$ did that one correctly and just defined int64 for everyone to use years ago before the C standards org decited that "long long" was legal.
Yah, this is really gotten on my nerves a number of times. I drive to Big Bend, or South Padre and on the way home they search my vehicle for Mexicans in my trunk. Its lots of fun if you have non American Tourists in your vehicle, especially when their visa's are about to expire.
Things like IE and media player arn't "integrated" with the OS. Just shipped with it. See what happens when you remove kthml or whatever the darn HTML rendere component in KDE is called now. All kinds of things stop working, help included same as windows. Just because you arn't techically savy enough to know how to remove indivual compoents in windows doesn't mean that other people arn't. Most people can't put diffrent seats in their cars either. This whole thing is truly amazing to me. The same people bitching about M$ shipping IE and media player are driving cars with CD players and radios integrated by the factory, and using Linux which not only usually comes with web browsers and media players but a host of other applications. How is a small commercial company making a web browser like opera suppost to compete with the likes of mozilla, konqueror, firefox etc.. on linux. Its the same thing. The only diffrence is that M$ spends all their energy making one good product instead of a dozen half ass ones and confusing their customers. (Ok thats not completly true, see access, sql server, and foxpro.. although they purchased 2 of the three.)
I concluded this about 3 years ago, when they started to try to avoid people gaming them. Back then I used to be able to type two very specific keywords (a OS platform, and the specific name of a piece of software I ported to that platform) into google and my page would appear. Now when I type those two keywords into google the "and" function doesn't seem to work, I get a lot of pages about the platform a few pages about the piece of software but nowhere is there any mention of my page where I maintain that piece of software for a particular platform. God only knows how many people would like to use my freely avialable software but can't find it because the "search" engines simply don't rank it high enough. The funny thing is that there are maybe a half dozen related pages that link to mine and the converse and we are all pretty much in a black whole 30 or 40 pages into the google rankings.
Of course if I type the whole title to my page I can get it but that is the point of a search engine, to figure out what you mean and display the appropriate page.
Yes, there is a legitimate argument here. The C++ and C standards committies move so slow that it probably will be another 10 years before they get another standard out. This of course is bad because it will probably take till 2010 to get the C changes into C++ but...
Anyway, there are other languages like this, fortran for example. All these little "kid" languages like VB, Java, C# etc.. are pretty much controlled by people who can just up and change the spec at a moments notice. The IEEE languages take years, and no one is willing to allow a competitor to break their stuff so it just sort of stalemates everything that has any possiblity of breaking anything. I've written a number of C++ programs abstracted through my own library call systems that are still being installed 15+ years after I wrote them, without any major vendor caused rewrites.
You also have the option to spend money on programmers capable of fixing any problems in the VB code/runtime to. I have done this in the past with just a good assembly language debugger and about 2 weeks of time. Its not impossible just damn hard. In truth this itself is a good reason to use a compiled language rather than an interpreted one. Its a lot easier to figure out the bugs in compiled code than figure out the bugs in some giant state machine running your code.
Slashdot Mirror
I think you need a better example, windows has since 3.0 (maybe longer) the standard where alt-command works on the global application space while ctrl-command works on the application space. Very few people know this, but its true. For example, Alt-f4 closes the application, ctrl-f4 closes the window. Alt-Tab cycles through applications, ctrl-tab cycles through the current applications windows.
I will agree with you though that as far as 'unix' goes OS-X is light years beyond any other 'unix' desktop. When i'm using linux one of the reasons I use KDE almost exclusivly is because they have tranditionally done a better job keeping application keystrokes consitant. While OS-X is _VERY_ nice any its pretty i'm not so sure about the keyboard usability. I can use a windows box without a mouse, i can't do that on a mac, maybe thats because i'm just not smart enough to know all the hotkeys.
Why would this be true? Can you see the difference between "strong" and "strongest"? In other words, that two strong competitors can thrive in the same marketplace?
Accually its simple, there are a number things that can happen.
There is a diffrence between investing in the military and spending more than the rest of the world put together...
Wake up and smell the stink in DC.
The Democrats and Republicans in this country are both the same with slight differences having nothing to do with conservatism or liberalism. The parties are more alike than different, and they are alike in the fact that they are heavily authoritarian .
Little out of touch are we? Windows is officially the largest server OS by installed base out there. It continues to grow pretty significalty by eating in the big unix servers market share. Linux is also growing at a much faster rate but they have a few years to catch up at the current growth rates, and they are also eating into the big unix servers market share. The number of people switching from windows->linux doesn't appear to be very large. All the growth is comming from unix->linux and unix->windows.
Proper application use of registry is to write local machine settings in HKLM and user settings into HKCU. Network based applications write to the servers HKLM and user specific settings to the HKCU. In fact the HKCU tree is pulled in over the network in a NT domain.
.INI files that people just like you used to bitch about because they got scattered all over the machine. Just like on linux, i have config files in /etc, /usr/local/etc, ~/, ~/.application, /var/something and a number of other places depending on which application we are talking about. Not that this is nessisary bad, except when application vendors put them in the wrong place. This is the same as the registry, when your running as administrator and you nice new application writes to HKLM when it should have written to HKLU its the same problem. You can find these applicatons in windows by running without permissions to update HKLM same as you can run linux/unix without permission to modify /etc/something.
Just because you don't understand the registry rules doesn't mean that its missing all these concepts you think it should have. As another poster pointed out the registry is an extension of the old
Its not really that simple, I tend to use VMware as a test and development bed. I've been purchasing MP machines for years, and usually my desktop machines have two processors. Anyway, with most things I use and write there is some level of MP scalablilty, vmware isn't any diffrent, when the virtual machine is idle it doesn't take up any physical CPU. With screen savers turned off (my virtual screen's don't need saving) and such the virtual machine will consume less than 1% of the physical CPU.
What really needs to be said is common knowledge with most people who work with MP machines, asside from some of the transaction based server machines, most programs have limits to how many concurrent operations they can run at the same time. Just like its hard to extract single thread instruction parralism past a couple of operations, its hard to extract multiple thread parralism past a couple of threads. My current application is scalable to somewhere between 5 to 10 threads, past that extra processors don't do anything for me. Basically the laws of software haven't changed in the last few years but the massive scalability of web based applications have helped people forget that throwing more processors at a problem even if its scalable won't always cause the application to scale.
Usually, somewhere someone has to consolidate all those transactions going to a DB, or whatever your bottleneck is. The attempt to sell MP machines to the masses is just a cheap and lazy way to speed peoples computers up. Problem sets that are processor bound and scalable are already running on big SMP's. Personally i like SMP's, but given the choice of a CPU 2x as fast or 2 cpu's I will take the single proc any day. Having two CPU's running 2x as fast is of course the best of both worlds and I'm usually happy to spend a little more to get that.
I predict that AMD, Intel etc will make a few multiproc CPU's get out to maybe 4 cores per die and then the race to make faster cores will kick back in. Only now people will expect screaming fast 4 way chips instead of screaming fast 1 way chips.
Everyone is saying "he must not be that smart/great" if he can't get a sound card working in Linux.. My take on it is "maybe he is to busy working on useful things to waste his time getting a sound card working in linux". As someone who gets shit done at home and work, I find that you have to focus on the problem and work on it rather than being distracted by everything around you. Spending 6 hours getting proper audio support in linux is 6 hours lost that could have been spent working on your project. Today it's 6 hours on the soundcard tomorrow its three weeks figuring out why the throughput in your application is 1/10th of what it should be (my current problem) because some idiot linux kernel 'hacker' broke part of the disk subsystem in the last 10 revisions of the kernel.
A few years ago I decided to switch my desktop back to windows 2k and exceed, and I'm significantly more productive than I ever was running linux, and wasting 3 hours trying to figure out how to remap my goto-line key in the most recent version of emacs, after the developers decided to use the key for something else.
This is silly, because while I may have 50 processes with 2 to 5 threads each, they are all idle waiting around for some event. The applications are all waiting for some network or user action, the "services" are all waiting around for network or applications to access them, the OS has a few house keeping threads eating up .01% of my CPU and the rest are running in serialized application contexts. In my case all the CPU time is being eaten up running virtual machines. Maybe vmware can burn up an extra CPU doing some kind of pre processing but I fail to see how a virtual machine will benifit from more CPU threads that are slower than one fast one...
This will just generate a secondary market in dive computers integrated with game boys.
Your just going to have to go back and get a cert for decompression diving, and learn to hang out decompressing
The other big drawback I see is that at depth the pressure of the water on your body is very great. That is why modern scuba uses pressure delivery systems.
I imagine that the air from this device will be generated near the pressure of the surrounding water, thereby simplifing the regulator.
It seems to me that appropriate selection of games would have a similar affect as that site. Warcraft or similar when played at full speed is going to increase your ability to make split second decisions and track lots of diffrent objects.
Back in the mid 90's my boss in response to me bitching about windows said "Its not that M$ writes good software, its that everyone else's software is worse" This was shortly after we dumped all our unix servers in favor of NT. Yes, at that time linux was around and we had even tried it for a while. I showed it running on one of my machines and my other boss managed to lock it up in under 3 minuites. The general consensus at the time where I worked? Linux was just another unix complete will all the usual unix problems.
Maybe now, but 10 years ago, a good reason to buy logitech was that they would ship a free lefty mouse with the purchase of a righty mouse. We did this at work a few times to justify the $50 mice. A righty would use the one that came in the package and a lefty would use the one that came in the mail. Two for 1...
Having a flashligh, crowbar and gloves in a car was enough to get a friend of mine taken to jail on some charge like felony posssion of buglary tools 10 years ago. Doesn't matter there was a toolbox full of other crap. The combination of the three was enough trigger some stupid law. He got out of it but thats not the point...
They can't figure out what they are doing. They stared out with a nice product, then instead of fixing it they have been playing palm works with "buzzword of month club" games.
I could give a fsck if its on a 300mhz ARM, i could give a fsck if its linux (which seems like a recipe for disaster on a handheld). Adding a non GUI thread scheduler should be easy( and not require a heavyweight OS like linux) As a developer what I wan't is a consistant API that extends rather than gets replaced every few years. Their product isn't even that good anymore. The new plams are more like crappy CE clones rather than Palm's which were always known for a unique and very stable system with loads of applications. Now days most of the applications seem like they are for palmOS 3.5 or 4, and the new ones don't scale well down to old devices. Plus palm is to busy making 100 diffrent versions of the same product to accually provide a nice range of expansion options. Then there are the tools, frankly the opensouce tools are a pain, it takes 1 day or so to get the enviroment working enough to write a hello world application. That is if your lucky or already know what to do. Writing a PalmOS 5 application with the opensource tools is a joke...
Now I just need to find a CE device with the battery life of my old Palm 505 and a similar footprint.
Cutting the fat is seen as acceptable because people assume that there is fat. That is accually a mangement failure too. It seems to indicate that there are a large number of people in the company that arn't doing their fair share of the work. This is probably true in a lot of these companies, but that accually stems from bad managment. If managment was accually keeping tabs on what everyone was doing they would know who just shows up and surfs /. all day instead of being productive. That situation would either be corrected or the person would be let go. Having people who were useful 10 years ago just sit around and surf is a bad idea for profitibility.
Hmmp. I would fly first class on a regular basis if the prices were more in line. Everytime I try to get a first class seat though it costs 5x as much as a normal seat for only about 2x the space.
Look at that table closer. Legacy mode includes "real-mode" and it does work in that mode. In "Long mode" it only supports 64-bit and 32-bit instructions. This is the mode that win64 runs in. If they wanted to run 16-bit applications they would have to either dynamically switch the processor back and forth between long and legacy, or write a software emulator to emulate the 16-bit instructions.
Accually most of the 64-bit unix's also have long=32 bits. I agree its stupid, long should have been 64-bits but now you have the "long long" basdardization which is 64-bits. At least M$ did that one correctly and just defined int64 for everyone to use years ago before the C standards org decited that "long long" was legal.
Yah, this is really gotten on my nerves a number of times. I drive to Big Bend, or South Padre and on the way home they search my vehicle for Mexicans in my trunk. Its lots of fun if you have non American Tourists in your vehicle, especially when their visa's are about to expire.
Things like IE and media player arn't "integrated" with the OS. Just shipped with it. See what happens when you remove kthml or whatever the darn HTML rendere component in KDE is called now. All kinds of things stop working, help included same as windows. Just because you arn't techically savy enough to know how to remove indivual compoents in windows doesn't mean that other people arn't. Most people can't put diffrent seats in their cars either. This whole thing is truly amazing to me. The same people bitching about M$ shipping IE and media player are driving cars with CD players and radios integrated by the factory, and using Linux which not only usually comes with web browsers and media players but a host of other applications. How is a small commercial company making a web browser like opera suppost to compete with the likes of mozilla, konqueror, firefox etc.. on linux. Its the same thing. The only diffrence is that M$ spends all their energy making one good product instead of a dozen half ass ones and confusing their customers. (Ok thats not completly true, see access, sql server, and foxpro.. although they purchased 2 of the three.)
I concluded this about 3 years ago, when they started to try to avoid people gaming them. Back then I used to be able to type two very specific keywords (a OS platform, and the specific name of a piece of software I ported to that platform) into google and my page would appear. Now when I type those two keywords into google the "and" function doesn't seem to work, I get a lot of pages about the platform a few pages about the piece of software but nowhere is there any mention of my page where I maintain that piece of software for a particular platform. God only knows how many people would like to use my freely avialable software but can't find it because the "search" engines simply don't rank it high enough. The funny thing is that there are maybe a half dozen related pages that link to mine and the converse and we are all pretty much in a black whole 30 or 40 pages into the google rankings.
Of course if I type the whole title to my page I can get it but that is the point of a search engine, to figure out what you mean and display the appropriate page.
Yes, there is a legitimate argument here. The C++ and C standards committies move so slow that it probably will be another 10 years before they get another standard out. This of course is bad because it will probably take till 2010 to get the C changes into C++ but...
Anyway, there are other languages like this, fortran for example. All these little "kid" languages like VB, Java, C# etc.. are pretty much controlled by people who can just up and change the spec at a moments notice. The IEEE languages take years, and no one is willing to allow a competitor to break their stuff so it just sort of stalemates everything that has any possiblity of breaking anything. I've written a number of C++ programs abstracted through my own library call systems that are still being installed 15+ years after I wrote them, without any major vendor caused rewrites.
You also have the option to spend money on programmers capable of fixing any problems in the VB code/runtime to. I have done this in the past with just a good assembly language debugger and about 2 weeks of time. Its not impossible just damn hard. In truth this itself is a good reason to use a compiled language rather than an interpreted one. Its a lot easier to figure out the bugs in compiled code than figure out the bugs in some giant state machine running your code.