Myself and everyone I work with has been doing this for many, many years, working on corporate class laptops, and using sleep/suspend 8-10 times per day, every business day.
And only very very rarely will a problem result from it. Vista is even better in that respect.
If you're seeing that kind of problems, then stop buying that type of computer.
Stick to the business class Dells (ie, Latitudes) and HP Compaq's. The drivers tend to be much, much better, which is often the problem that causes what you describe.
Thats one of the reasons you dont ever want to buy consumer garbage like Toshiba, Sony, etc. Stick with corporate class equipment from Dell, HP or IBM (not sure how Lenovo is nowadays) if your computer is an essential tool for your work.
- much better desktop experience due to the compositing window manager, no tearing, etc.
- MUCH better I/O scheduler, which makes the machine usable even when the drive is getting hammered
- much better start-menu/search... just hit the windows key, type the first couple letters of what you want to start, and hit return
- much higher security configuration with UAC (though I mean mostly in the way you can configure, rather than the way it comes configured)
- layered integrity model, which prevents shatter type attacks and a whole other host of attacks
- hugely improved 64-bit support
- a ton more drivers in userland (which increases reliability)
etc etc.
It's not perfect, and I swear Windows Explorer gets worse with every release of windows, and the networking configuration UIs are freaking horrible in Vista. But for the business user its got alot of nice improvements that just makes it more reliable and predictable.
That's far from trivial. That undercuts any notion that object-oriented programming is intended to model the problem at hand.
It sounds like you're assuming that there should only be one domain/concern to model, but the real world never works like that. You always have competing conceptual models (or at least pieces of them) in the business domain, and you always have real-life technical artifacts/limitations that influence your models. That doesnt mean OO is useless, it just means that absolute purity is difficult in the real world.
And tweaking the way you design systems to support unit testing tends to actually improve most developers structures/factoring, in my experience. It's a beautiful side effect that seems to work quite serendipitously in reality.
Unless you're doing something on the scale of an operating system, what could possibly warrant three million lines of code in one program? In eleven years programming professionally, the largest program I've ever written was ~10kloc (and I eventually regretted how monolithic I made it) and the largest I've worked on (awful monolithic and repetitive architecture) was around ~50kloc.
An enterprise financial system for large national (but not multi-national at this point) orgs. Including procurement/purchasing, reporting, workflow, batch system, authn/authz frameworks, all done in a web app. It's an ERP, they're just huge, at least from what I've seen.
Note also that this is done in Java, which is a rather verbose language.:) So that inflates the numbers a little bit.
That's my impression, but people who actually use scripting languages for a living (I do.Net and PL/SQL lately) seem to dispute it.
I do both... we also do some Rails contract dev recently. Some things I love about Ruby & Rails, but the tooling is freaking terrible. At least in some ways. If you consider the rails generator and rake as tooling, then that helps, but is not any more fundamentally powerful (though much more concise) than hibernate.
Ruby has one, anyway. It's not pretty, but having an interactive interpreter tied to the debugger is handy; that's a big part of why people could get stuff done in VB6 despite the piss-poor language, and Ruby's is more versatile.
Ruby definitely does not unless one just showed up in the past year or so. You can add a line of code that acts as a break point in the system, and then fall out to the interactive ruby prompt. But thats not even close to the same thing as the interactive debugger I have with Eclipse in Java, or hell, even close to as powerful as the VB6 debugger I had 10 years ago.
If I am missing something fundamental, and there is a real stepping debugger (ie, let me step into, step over, step out, and see the full stack trace at any point the value of any variables), PLEASE let me know.
testability just doesn't fit very well into the Java/.Net paradigm.
I dont understand this at all.
Like any language, in Java or.NET, if you're doing unit testing, it changes the way you design the objects and structures. You design them to BOTH fit the domain model and to more easily support unit testing. But there's nothing inherently difficult about doing unit testing and a TDD approach with Java or.NET.
Now, granted, a language could let you declare "MyClass myObject", validate it, and still let you change what those mean for testing; that would be great. Until languages like that appear, though, I think I'm sold on Python and Ruby being more reliable for practical purposes than Java and.Net.
Thats whats so great about static typed languages, you have Interfaces. Combine that with dependency injection and mocks, and your whole last two paragraphs are trivially answered.
It's true that its a bit more overhead and setup for very small apps, but once you get past a few thousand or tens of thousands of lines of code, I would argue it makes it more manageable.
I think one of the big things statically typed languages gives you is better tooling. One of my contracts right now is a ~3M loc system in Java, which is a heavy user of spring and hibernate. I can move around this thing so fast, follow code execution paths, call & type hierarchies, etc. If I need to rename a method, or add/remove parameters, the IDE immediately tells me what it breaks, or does the fixups for me automatically using the refactoring tools.
And I have a stepping debugger! Why the heck cant any modern dynamically typed languages include a debugger. I realize that a debugger isnt for every day use. But when you've got to drop into a complex set of code written by someone else and step through a few hundred calls to figure out why something is broken, a stepping debugger is essential. Without it, you have to litter your code with log dumps and other noise.
Lastly, at least for languages like Python & Ruby, it fundamentally breaks one of the core concepts of OO: Encapsulation. In these languages you're FORCED to interact with other components based on your looking at the source (ie, their implementation) rather than their published interface. That is just so broken its hard to comprehend why no one else complains about it.
I see you didnt notice my second paragraph. Running that size of a load in 4 thumpers with gigantic 1TB or bigger drives is not reasonable. These are machines in which every spindle is hammered 24x7, and you are completely I/O bottlenecked at all times, against those drives.
As you move to bigger drives, the number of concurrent users queued up against reads & writes on a given spindle goes way up. Plus the pain/cost/downtime when each drive fails goes way up, as you have so very many mailboxes on each drive.
What would you say it costs for people, even if we assume your non-practical big-drive solution? Are you seriously suggesting that its zero? You need to include the costs for purchasing people (since that size of a purchase requires an RFP unless you already have an exclusive contract at a university), the setup, installation, migration, support techs to deal with end-users when the problems happen. Plus in addition to the low end techs who do some of the work, you need some actually good people (which are very rare in higher ed) to deal with the hard problems.
Then you have to deal with tape backup systems (which usually cost 2-3x the storage cost per unit), and increased power cooling and bandwidth load in both your primary and secondary datacenter.
Regarding your startup company comment, you're right! Running a startup IS different than running email services for 30-50k users in a University environment. That should not be a surprise. I dont see how its germane to the conversation though.
For what its worth, I do own my own business, and as stated above, I do run it differently than I ran IT at large organizations. This should also not be a surprise to anyone.
Wow. Please trust me when I say that this would be a terrible, terrible decision.
Just the storage, backup, disaster-recovery, and bandwidth cost to support 30-50k students, and the associated management costs, assuming all open-source software, will be mid 6-figures yearly.
You also seem to be confusing an SMTP server with email storage, mailboxes, spam & virus protection, backups, etc etc.
Thats a nice attempt to work it out, but it completely misses what is often the biggest cost: PEOPLE.
Will those drives install themselves? The project has to be planned, equipment needs to be purchased, communications need to happen, downtime has to be scheduled, new equipment installed, tested, and then dealt with when there are problems. The larger systems require an incremental increase in staff to manage.
Not to mention the rank insanity of running 1TB+ spindles in a system like that which is pure disk I/O limited. It'll be slow, and you'll take out alot of people when spindles fail. Thats not how you design real systems.
I think you are having two account books here. I will accept your numbers (why I shouldn't?) but I can't accept your conclussion: even if your higher quota costs US$500.000, that's a one shoot cost for (probably) between three and five fiscal years; that's peanuts.
That is NOT a one-shot cost.
Typically, for every $1 that direct storage costs you, backups cost you between $2-3 for that same unit. And that is NOT a one time cost.
As the amount of data you handle goes up, your management cost goes up too. You hope that you achieve economies of scale when you grow, and the percentage of what it costs you to manage goes down, but the actual dollar cost still goes up.
And $500k is NOT peanuts, even to a large state University. Especially when state funding is going down across the country every year.
Many Universities are seeing budgets shrinking every year right now. In that scenario $500k is not trivial.
Even accepting that you will save 250.000US$/year (which I have problems to accept without see the detailed numbers) that's not a for-profit company, that's a f* University.
For a University with 50k active accounts (so figure 30-50k students) 6-figures per year to run an email in-house is very reasonable. Thats not an outrageous number.
Going to University (and maintaining them) is more expensive that not going or having them, at least on the short term, isn't it? While not the main point, one of the very points of having universities (and/or technical colleges) is in order to "try to scale Cyrus"; it's not per chance that Cyrus itself comes from a University.
Building IT experience or testing an open-source mail product is utterly and completely NOT the point of having a University. Now you may get a grant to do one of those things, and run it through the University, and thats fine. But the state or endowment funds that go to run the things needs to be spent first and foremost on teaching students.
Your priorities are completely out of whack with reality. Universities run on a zero-sum economic situation. So which classes should be cancelled or faculty not be hired to run that mail system? Thats the choice. It's not just 'the tax system will fund it'. Thats not how it works.
CS people dont do work like this. You're thinking of IT.
And its all about the money. At least in the US, there is a frightening phenomenon of government funding for higher ed dramatically falling off. So Universities struggle in how to provide services, particularly ones that arent core.
Besides, providing email services to 30-50k students, with a 5-8k turnover each year is not exactly a simple task.
It costs a lot of money (6-7 figures US yearly) to run it in-house at a University that size.
Most CS departments, at least in large Universities that I've seen, do NOT have their own datacenters. They may host some things in the central data centers, or may have some small 'server rooms' for 10 or 20 servers, or similar.
If it's so trivial that any application run under RunAs behaves exactly as if run after logging in interactively as the same user, why do you have to "understand how the system works"?
The applications/processes dont have to know or care anything about it. The end-user sometimes does because that one process is running under a different account than the desktop. So if that RunAs'd process makes changes to the profile, the end-user will have to understand that it's making changes to the RunAs'd user account profile, NOT the desktop profile. If you want to do that, then you use MakeMeAdmin or similar tools.
I really don't know how RunAs works, but if you think about sudo on unix, it's not quite as simple as that. If you've originally logged in as "foo" and then run a command as user "bar" through sudo, does the application you run get the environment (environment variables such as locale settings, home directory location, etc.) from foo or from bar? Or a part from one and another part from the other.
RunAs isnt equivalent to sudo. It's equivalent to su. And just like su, you can choose whether or not to load the whole profile.
Vista's UAC is more like sudo, at least when run in certain configurations, just not nearly as configurable.
Of course you can make it behave they way that is appropriate for the situation. Perhaps it's the same with RunAs and what you mean by "not understanding" it is exactly that. But "zero difference" between RunAs and an interactive user session may still not quite hold due to such details, and the existence of a magical flag to for applications to detect RunAs is likewise irrelevant.
What I mean is that from the process' (ie, the executable running) perspective, there is zero difference between being RunAs'd as JoeAdmin and running under the JoeAdmin desktop. There can be secondary differences if the process makes changes to the profile, and the end-user doesnt understand that its running under a completely different profile.
So from the process/executable perspective, there is no difference. It doesnt care, and doesnt have to be programmed any differently to take it into account.
From the end-user's perspective, there is a difference, particularly if the thing you want to RunAs changes the user profile.
The latter. It's currently available through mid 2009, and looks like it'll be extended further if there is customer demand.
I have no idea whether small white-box builders (ie, system builders) can use the downgrade program, but I cant imagine why not. No personal experience, in any case.
As mentioned in previous posts, there are ways for our customers to voluntarily and anonymously send us data on how they use our features (Customer Experience Improvement Program, Windows Feedback Panel, user surveys, user in field testing, blog posts, and in house usability testing). The data and feedback we collect help inform and prioritize the decisions we make about our feature designs. From this data, weve learned a lot about UACs impact.
Customer Experience Improvement Program data indicates that the number of sessions with one or more UAC prompts has declined from 50% to 33% of sessions with Vista SP1.
So configure UAC to behave the way you want. It takes about 90 seconds.
Configure it so that Admin accounts dont require elevation at all, but non-admin accounts get prompted for elevation credentials (ie, ask for a user & pass).
Thats trivial to show how it makes Vista more secure.
On Vista, when running in home-user mode, with admin approval, you get a prompt when something wants to make system level changes.
On XP, when running in home-user mode, as default, you get no warning. The changes just happen.
Clearly and obviously Vista is the more secure situation. You get warning, and you can say no. In XP, you neither got warning nor did you get to say no.
Thats because windows ships by default configured for home users, who literally are generally incapable of dealing with remembering passwords.
Businesses dont leave it configured like that. They make all users run as non-admin and elevate with a user/password. It's just a minor configuration change in UAC.
I would argue that you may not be a very good admin, based on the information you wrote.
Do you ever use web browsers? Do you ever browse to any websites, even perfectly safe ones like Amazon.com?
If so, there are numerous drive-by-installers, even on Firefox, even on Opera. All you have to do is go to a website. Even mainstream ones, they get owned all the time too. Or their advertising providers do.
If this happens and you browse to amazon.com and you are running as admin, you can get owned instantly. If your'e not running as admin, this will have no effect.
I'm aware of the issue but all those prompts aren't for backward compatibility. They're a way of MS shifting the security burden on the dumb users.
Not quite. It's a way to force the burden to bad app developers, and its working. It's a way to force the ecosystem to clean itself up.
MS had to do it eventually, and it was never going to be easy. The windows software ecosystem will be a better place because they did this, at the cost of some pain now.
Slashdot Mirror
Myself and everyone I work with has been doing this for many, many years, working on corporate class laptops, and using sleep/suspend 8-10 times per day, every business day.
And only very very rarely will a problem result from it. Vista is even better in that respect.
If you're seeing that kind of problems, then stop buying that type of computer.
Stick to the business class Dells (ie, Latitudes) and HP Compaq's. The drivers tend to be much, much better, which is often the problem that causes what you describe.
Thats one of the reasons you dont ever want to buy consumer garbage like Toshiba, Sony, etc. Stick with corporate class equipment from Dell, HP or IBM (not sure how Lenovo is nowadays) if your computer is an essential tool for your work.
-
Add /SOS to the boot line.
Gives you mostly what you want. But to be honest, it goes by so fast its almost unusable.
Even the linux based ones from DirectTV do that.
My POS drops a few seconds right on the 15th minute of nearly every 30-minute show I record.
That thing is so damn flaky.
Even on an airplane, use standby, dont actually shut the thing off.
There's just a trickle of current to keep the ram alive, its no problem.
Been flying that way for years, with nary a problem.
What about:
- massively improved reliability over XP
- much better desktop experience due to the compositing window manager, no tearing, etc.
- MUCH better I/O scheduler, which makes the machine usable even when the drive is getting hammered
- much better start-menu/search ... just hit the windows key, type the first couple letters of what you want to start, and hit return
- much higher security configuration with UAC (though I mean mostly in the way you can configure, rather than the way it comes configured)
- layered integrity model, which prevents shatter type attacks and a whole other host of attacks
- hugely improved 64-bit support
- a ton more drivers in userland (which increases reliability)
etc etc.
It's not perfect, and I swear Windows Explorer gets worse with every release of windows, and the networking configuration UIs are freaking horrible in Vista. But for the business user its got alot of nice improvements that just makes it more reliable and predictable.
That's far from trivial. That undercuts any notion that object-oriented programming is intended to model the problem at hand.
It sounds like you're assuming that there should only be one domain/concern to model, but the real world never works like that. You always have competing conceptual models (or at least pieces of them) in the business domain, and you always have real-life technical artifacts/limitations that influence your models. That doesnt mean OO is useless, it just means that absolute purity is difficult in the real world.
And tweaking the way you design systems to support unit testing tends to actually improve most developers structures/factoring, in my experience. It's a beautiful side effect that seems to work quite serendipitously in reality.
Unless you're doing something on the scale of an operating system, what could possibly warrant three million lines of code in one program? In eleven years programming professionally, the largest program I've ever written was ~10kloc (and I eventually regretted how monolithic I made it) and the largest I've worked on (awful monolithic and repetitive architecture) was around ~50kloc.
An enterprise financial system for large national (but not multi-national at this point) orgs. Including procurement/purchasing, reporting, workflow, batch system, authn/authz frameworks, all done in a web app. It's an ERP, they're just huge, at least from what I've seen.
Note also that this is done in Java, which is a rather verbose language. :) So that inflates the numbers a little bit.
That's my impression, but people who actually use scripting languages for a living (I do .Net and PL/SQL lately) seem to dispute it.
I do both ... we also do some Rails contract dev recently. Some things I love about Ruby & Rails, but the tooling is freaking terrible. At least in some ways. If you consider the rails generator and rake as tooling, then that helps, but is not any more fundamentally powerful (though much more concise) than hibernate.
Ruby has one, anyway. It's not pretty, but having an interactive interpreter tied to the debugger is handy; that's a big part of why people could get stuff done in VB6 despite the piss-poor language, and Ruby's is more versatile.
Ruby definitely does not unless one just showed up in the past year or so. You can add a line of code that acts as a break point in the system, and then fall out to the interactive ruby prompt. But thats not even close to the same thing as the interactive debugger I have with Eclipse in Java, or hell, even close to as powerful as the VB6 debugger I had 10 years ago.
If I am missing something fundamental, and there is a real stepping debugger (ie, let me step into, step over, step out, and see the full stack trace at any point the value of any variables), PLEASE let me know.
testability just doesn't fit very well into the Java/.Net paradigm.
I dont understand this at all.
Like any language, in Java or .NET, if you're doing unit testing, it changes the way you design the objects and structures. You design them to BOTH fit the domain model and to more easily support unit testing. But there's nothing inherently difficult about doing unit testing and a TDD approach with Java or .NET.
Now, granted, a language could let you declare "MyClass myObject", validate it, and still let you change what those mean for testing; that would be great. Until languages like that appear, though, I think I'm sold on Python and Ruby being more reliable for practical purposes than Java and .Net.
Thats whats so great about static typed languages, you have Interfaces. Combine that with dependency injection and mocks, and your whole last two paragraphs are trivially answered.
It's true that its a bit more overhead and setup for very small apps, but once you get past a few thousand or tens of thousands of lines of code, I would argue it makes it more manageable.
I think one of the big things statically typed languages gives you is better tooling. One of my contracts right now is a ~3M loc system in Java, which is a heavy user of spring and hibernate. I can move around this thing so fast, follow code execution paths, call & type hierarchies, etc. If I need to rename a method, or add/remove parameters, the IDE immediately tells me what it breaks, or does the fixups for me automatically using the refactoring tools.
And I have a stepping debugger! Why the heck cant any modern dynamically typed languages include a debugger. I realize that a debugger isnt for every day use. But when you've got to drop into a complex set of code written by someone else and step through a few hundred calls to figure out why something is broken, a stepping debugger is essential. Without it, you have to litter your code with log dumps and other noise.
Lastly, at least for languages like Python & Ruby, it fundamentally breaks one of the core concepts of OO: Encapsulation. In these languages you're FORCED to interact with other components based on your looking at the source (ie, their implementation) rather than their published interface. That is just so broken its hard to comprehend why no one else complains about it.
As a minor correction, only the HTML rendering component of AIR is built on WebKit.
Thats a minor (albeit useful) piece of AIR.
I see you didnt notice my second paragraph. Running that size of a load in 4 thumpers with gigantic 1TB or bigger drives is not reasonable. These are machines in which every spindle is hammered 24x7, and you are completely I/O bottlenecked at all times, against those drives.
As you move to bigger drives, the number of concurrent users queued up against reads & writes on a given spindle goes way up. Plus the pain/cost/downtime when each drive fails goes way up, as you have so very many mailboxes on each drive.
What would you say it costs for people, even if we assume your non-practical big-drive solution? Are you seriously suggesting that its zero? You need to include the costs for purchasing people (since that size of a purchase requires an RFP unless you already have an exclusive contract at a university), the setup, installation, migration, support techs to deal with end-users when the problems happen. Plus in addition to the low end techs who do some of the work, you need some actually good people (which are very rare in higher ed) to deal with the hard problems.
Then you have to deal with tape backup systems (which usually cost 2-3x the storage cost per unit), and increased power cooling and bandwidth load in both your primary and secondary datacenter.
Regarding your startup company comment, you're right! Running a startup IS different than running email services for 30-50k users in a University environment. That should not be a surprise. I dont see how its germane to the conversation though.
For what its worth, I do own my own business, and as stated above, I do run it differently than I ran IT at large organizations. This should also not be a surprise to anyone.
Wow. Please trust me when I say that this would be a terrible, terrible decision.
Just the storage, backup, disaster-recovery, and bandwidth cost to support 30-50k students, and the associated management costs, assuming all open-source software, will be mid 6-figures yearly.
You also seem to be confusing an SMTP server with email storage, mailboxes, spam & virus protection, backups, etc etc.
Thats a nice attempt to work it out, but it completely misses what is often the biggest cost: PEOPLE.
Will those drives install themselves? The project has to be planned, equipment needs to be purchased, communications need to happen, downtime has to be scheduled, new equipment installed, tested, and then dealt with when there are problems. The larger systems require an incremental increase in staff to manage.
Not to mention the rank insanity of running 1TB+ spindles in a system like that which is pure disk I/O limited. It'll be slow, and you'll take out alot of people when spindles fail. Thats not how you design real systems.
I think you are having two account books here. I will accept your numbers (why I shouldn't?) but I can't accept your conclussion: even if your higher quota costs US$500.000, that's a one shoot cost for (probably) between three and five fiscal years; that's peanuts.
That is NOT a one-shot cost.
Typically, for every $1 that direct storage costs you, backups cost you between $2-3 for that same unit. And that is NOT a one time cost.
As the amount of data you handle goes up, your management cost goes up too. You hope that you achieve economies of scale when you grow, and the percentage of what it costs you to manage goes down, but the actual dollar cost still goes up.
And $500k is NOT peanuts, even to a large state University. Especially when state funding is going down across the country every year.
Many Universities are seeing budgets shrinking every year right now. In that scenario $500k is not trivial.
Even accepting that you will save 250.000US$/year (which I have problems to accept without see the detailed numbers) that's not a for-profit company, that's a f* University.
For a University with 50k active accounts (so figure 30-50k students) 6-figures per year to run an email in-house is very reasonable. Thats not an outrageous number.
Going to University (and maintaining them) is more expensive that not going or having them, at least on the short term, isn't it? While not the main point, one of the very points of having universities (and/or technical colleges) is in order to "try to scale Cyrus"; it's not per chance that Cyrus itself comes from a University.
Building IT experience or testing an open-source mail product is utterly and completely NOT the point of having a University. Now you may get a grant to do one of those things, and run it through the University, and thats fine. But the state or endowment funds that go to run the things needs to be spent first and foremost on teaching students.
Your priorities are completely out of whack with reality. Universities run on a zero-sum economic situation. So which classes should be cancelled or faculty not be hired to run that mail system? Thats the choice. It's not just 'the tax system will fund it'. Thats not how it works.
CS people dont do work like this. You're thinking of IT.
And its all about the money. At least in the US, there is a frightening phenomenon of government funding for higher ed dramatically falling off. So Universities struggle in how to provide services, particularly ones that arent core.
Besides, providing email services to 30-50k students, with a 5-8k turnover each year is not exactly a simple task.
It costs a lot of money (6-7 figures US yearly) to run it in-house at a University that size.
Most CS departments, at least in large Universities that I've seen, do NOT have their own datacenters. They may host some things in the central data centers, or may have some small 'server rooms' for 10 or 20 servers, or similar.
If it's so trivial that any application run under RunAs behaves exactly as if run after logging in interactively as the same user, why do you have to "understand how the system works"?
The applications/processes dont have to know or care anything about it. The end-user sometimes does because that one process is running under a different account than the desktop. So if that RunAs'd process makes changes to the profile, the end-user will have to understand that it's making changes to the RunAs'd user account profile, NOT the desktop profile. If you want to do that, then you use MakeMeAdmin or similar tools.
I really don't know how RunAs works, but if you think about sudo on unix, it's not quite as simple as that. If you've originally logged in as "foo" and then run a command as user "bar" through sudo, does the application you run get the environment (environment variables such as locale settings, home directory location, etc.) from foo or from bar? Or a part from one and another part from the other.
RunAs isnt equivalent to sudo. It's equivalent to su. And just like su, you can choose whether or not to load the whole profile.
Vista's UAC is more like sudo, at least when run in certain configurations, just not nearly as configurable.
Of course you can make it behave they way that is appropriate for the situation. Perhaps it's the same with RunAs and what you mean by "not understanding" it is exactly that. But "zero difference" between RunAs and an interactive user session may still not quite hold due to such details, and the existence of a magical flag to for applications to detect RunAs is likewise irrelevant.
What I mean is that from the process' (ie, the executable running) perspective, there is zero difference between being RunAs'd as JoeAdmin and running under the JoeAdmin desktop. There can be secondary differences if the process makes changes to the profile, and the end-user doesnt understand that its running under a completely different profile.
So from the process/executable perspective, there is no difference. It doesnt care, and doesnt have to be programmed any differently to take it into account.
From the end-user's perspective, there is a difference, particularly if the thing you want to RunAs changes the user profile.
The latter. It's currently available through mid 2009, and looks like it'll be extended further if there is customer demand.
I have no idea whether small white-box builders (ie, system builders) can use the downgrade program, but I cant imagine why not. No personal experience, in any case.
From TFA:
As mentioned in previous posts, there are ways for our customers to voluntarily and anonymously send us data on how they use our features (Customer Experience Improvement Program, Windows Feedback Panel, user surveys, user in field testing, blog posts, and in house usability testing). The data and feedback we collect help inform and prioritize the decisions we make about our feature designs. From this data, weve learned a lot about UACs impact.
Customer Experience Improvement Program data indicates that the number of sessions with one or more UAC prompts has declined from 50% to 33% of sessions with Vista SP1.
So configure UAC to behave the way you want. It takes about 90 seconds.
Configure it so that Admin accounts dont require elevation at all, but non-admin accounts get prompted for elevation credentials (ie, ask for a user & pass).
Thats utterly trivial to do on Vista.
In IIS 7, Microsoft replaced the ADSI with XML files.
Nearly everything about this sentence is wrong.
IIS6 (not 7) was when IIS started using XML files for configuration.
IIS NEVER stored configuration in 'ADSI'. ADSI = Active Directory Services Interface, and has nothing to do with IIS.
Thats trivial to show how it makes Vista more secure.
On Vista, when running in home-user mode, with admin approval, you get a prompt when something wants to make system level changes.
On XP, when running in home-user mode, as default, you get no warning. The changes just happen.
Clearly and obviously Vista is the more secure situation. You get warning, and you can say no. In XP, you neither got warning nor did you get to say no.
How is that not clear?
Thats because windows ships by default configured for home users, who literally are generally incapable of dealing with remembering passwords.
Businesses dont leave it configured like that. They make all users run as non-admin and elevate with a user/password. It's just a minor configuration change in UAC.
I would argue that you may not be a very good admin, based on the information you wrote.
Do you ever use web browsers? Do you ever browse to any websites, even perfectly safe ones like Amazon.com?
If so, there are numerous drive-by-installers, even on Firefox, even on Opera. All you have to do is go to a website. Even mainstream ones, they get owned all the time too. Or their advertising providers do.
If this happens and you browse to amazon.com and you are running as admin, you can get owned instantly. If your'e not running as admin, this will have no effect.
I'm aware of the issue but all those prompts aren't for backward compatibility. They're a way of MS shifting the security burden on the dumb users.
Not quite. It's a way to force the burden to bad app developers, and its working. It's a way to force the ecosystem to clean itself up.
MS had to do it eventually, and it was never going to be easy. The windows software ecosystem will be a better place because they did this, at the cost of some pain now.
You need to update your information.
Go to any major OEM, you can still buy machines with XP.
Yes. It's called application compatibility shims or aplication virtualization.
Thats why they included it with Vista by default.
However, in the real world it doesnt ever work out as simply as you describe, which is why it doesnt take over and virtualize everything by default.
Read up on it, what you describe is already in Vista.