with the internet, u're confined to (a) the size of an IP packet, and (b) ur pipe.
It's no worse than telnetting to a remote server and running a keystroke interactive app (e.g. emacs). And I can confirm that that works fine, even on a 14.4k modem.
In my IT consultancy business, we used to use Linux desktops for a time. When we did this, we would use ytalk as a primitive IM application; it was primarily useful if one of us needed to get information from another while talking to a client on the phone. It helped present an aura of "these people know what they're doing."
We switched to Windows desktops when it became apparent that we needed to use Internet Explorer for compatibility testing on too regular a basis for any other solution, and at the time there was no obvious replacement for ytalk. We eventually started using modern IM software when it became widespread, but no longer find as much need for it.
I suppose I could further increase my speed by relying on shorthand, but I can't bring myself to butcher the language like that (eg. c u l8r!)
With predictive entry, it's actually harder to write that. On my Nokia, "c u l8r" takes 14 button presses in predictive mode, or 17 in non-predictive, whereas "see you later" takes only 13 -- plus it's much faster because getting the '8' requires going through a menu which involves waiting while the menu is displayed.
I don't think so. The CPU is spending most of its time idle on most machines, so why do drivers for SLOW HARDWARE have to be running at kernel speed?
Because that would make the hardware slower, and it is because the hardware is slower than the CPU that the CPU is spending most of its time idle (i.e. waiting for hardware to finish its job before it can get on with the next one).
Having debugged a network driver problem on one of my machines lately, I can tell you that a couple of milliseconds extra latency in responding to an IRQ from a network card can drop the maximum throughput of an SMB file server from ~50% utilisation down to ~5%. That's a huge difference. Extrapolating to a more reasonable delay of a few hundredths of a millisecond and you'll still see a noticeable performance degradation.
You'll notice it more with faster hardware. Add a hundredth of a millisecond delay to each hard disk access and you'll see file system performance drop by a significant figure, maybe as much as 10%.
Sound card -- that probably won't make much difference, except for professional recording studio applications. But they need an OS too, so we might as well keep the driver in there for them.
Latency's a killer. Doing everything you can to reduce it really increases throughput.
I figure, you'd want to replace something like WINLOGON.EXE, or whatever the closest equivalent of init there is on Windows. I'm sure there are people here who are a lot more knowledgable about how WinNT starts.
I suspect you'd have to replace either CSRSS.EXE or SMSS.EXE, and the app you replace it with would have to be a native application, so it couldn't be CMD.EXE which is a win32 console subsystem application. More info on sysinternals, here and here.
Note that I/O will be your primary difficulty -- the only API available to you was designed for output only during the blue screen phase of Windows NT's boot process, and for display BSODs. You will probably have to install a device driver that enables access to a text console and use that for IO.
This can be done, as both Windows Setup and the Recovery Console seem to use this approach.
OK -- I'm running Pro and get task manager directly from ctrl-alt-del. How do I get to the other screen you're talking about? (I used to find it useful on W2K)
Wireless Zero Configuration (if you have a wireless card you should probably use the vendor's config interface -- the Windows one sends a query out for avaiable networks every minute which causes delays with some cards)
Secondary Logon, Terminal Services, Fast User Switching Compatibility (these are needed if you want to use fast user switching, but many people don't, so you can safely disable them in this case. You can disable Fast User Switching Compatibility anyway -- I've not found anything that needs it.)
Computer Browser (doesn't seem to do anything useful) Server, Workstation (unless you need to share files or printers or use such shares respectively) Cryptographic Services Application Layer Gateway
Err... the firewall that most people use is the one in the "Windows Firewall/Internet Connection Sharing" service. I've never heard of anybody who actually does what's suggested in that link.
If you want something lighter weight than that, remove all network services other than TCP/IP from your network adapter and then enable the port-based filtering on the TCP/IP Advanced Settings dialog. You can't do source based filtering, but few people actually need to do that.
I run XP on a machine with only 96Mb of RAM. Needless to say, I've experimented a lot with which services I actually need to be running. The ones I still have are:
Automatic Updates (couldn't live without it, unfortunately) COM+ Event System (some of the management processes use it) DHCP Client (I could assign myself a permanent IP address, but I'm too lazy to figure out how to do that with my router) Event Log (cause it's kind of handy sometimes) Plug & Play RPC (cause other stuff depends on it) Server (I use it, you might not...) Shell Hardware Detection System Event Notification (I think I need this, I'm not entirely sure though) TCP/IP NetBIOS Helper Windows Audio Workstation
However, the reason Firefox is slower is because it has the XPCOM-platform-abstraction-layer and uses the Javascript-bindings for core-functionality (browser.js is the actual browser; I'm not joking), which eases cross-platform development, but causes performance-penalties.
Were Opera not a cross platform browser, I'd agree that you have a point there. But, Opera have clearly shown that you can have a working browser on multiple platforms without it being slow.
My personal opinion is that too much of Firefox is implemented in JS. Probably half of that JS could be easily reimplemented as native code for a substantial speed bonus. I'd like to see that browser.
I personally have noticed no speed difference [between IE and Firefox], but I have a fast machine.
For reference, on my PII-400 I'd say firefox takes about 2-3x as long to start up, and frequently suffers long delays in various actions. Particularly grievous is the long (~200ms) pause that frequently occurs after typing the second letter of a URL in the address box while it looks up history items starting with those two letters. This pause is also noticeable on a Celeron 1.3GHz laptop, although nothing like as annoying.
Firefox also seems to use about 50% more memory on average for the same operation. It is also noticeable that it only uses single threads for many things where IE uses multiple: if one window starts a plugin, for example, all the others freeze until after the plugin has finished initialising.
Thunderbird is worst -- my entire machine grinds to a halt while it displays the new message notification window.
Even if you do notice a difference, any semi-intelligent human being knows that a 10% increase in speed isn't everything. Firefox has so much more to offer.
True, and that's why I continue to use it, despite the inconvenience. I wouldn't give up tabbed browsing for anything, for instance.
I'll be giving IE7 a try once it comes out of beta.
You're talking about the triangles in the navigation structures, right? Why do you think this is some kind of KDE innovation? I've been using images almost identical to those in web site navigation systems since some time around 1998[1]. It's hard to think of a simpler set of symbols to suggest the concepts we're talking about here. This is exactly why patents on user interfaces are so absurd -- usually any individual idea is so obvious it will be implemented hundreds of times by people who have no idea about what each other are doing.
[1] I don't know when KDE started using them like that, but I hadn't ever seen or used KDE at the time.
Yes, although it's not quite as simple as the OP said. There is a defense of not having the key: you just have to be able to provide enough evidence that you might not that it calls the prosecution's evidence that you do into doubt.
the RIP Act isn't available to enough authorities, or is too hard to invoke or something.
Err, no, not really. It's actually too easy to invoke -- a simple piece of paper is all that's required, reading something like this:
NOTE OF REQUIREMENT TO DISCLOSE ENCRYPTION KEY
You are hereby required under the Regulation of Investigatory Powers Act 2000 to provide copies of any and all encryption keys to which you have access which are assocaited with the email address suspect@example.com. You are furthermore ordered not to disclose the contents of this order to any individual except authorised representatives of the police force, and your legal adviser if any. You have until 10am tomorrow morning to comply.
Signed
(A chief police officer, a police officer authorised to sign on behalf of a chief police officer, a justice of the peace, or similar).
No. What happens is you get a notice and that notice tells you that you aren't allowed to reveal the existence of the notice to anyone else. Which means in your example, doing anything other than answering "no" might be a breach of the requirement.
If you actually look at the act you'll see it says:
For the purposes of this section a person shall be taken to have shown that he was not in possession of a key to protected information at a particular time if-
(a) sufficient evidence of that fact is adduced to raise an issue with respect to it; and
(b) the contrary is not proved beyond a reasonable doubt.
That's not proof beyond a reasonable doubt, that's enough evidence to suggest it might be the case and it can't be proved beyond reasonable doubt the other way.
but what "400% efficiency" means in the rest of the world is "takes one unit of energy as input, and outputs four units of energy" -- i.e. free energy.
You can of course use this technique to get "free energy" (i.e. energy captured from the environment). See heating systems currently sold as "geothermal" (which aren't reallygeothermal, they just capture heat from the ground that has been warmed up by the sun).
A random thought: I thought iTunes didn't sell Beatles music, due to the trademark issues surrounding the "Apple" name related to music? Has this now been resolved?
Windows: good for getting those right-click menus. Also the only way to do things that don't have obvious keyboard shortcuts - preference dialogs, toolbar buttons, etc.
I'm a windows users who has an instinct to use the keyboard. Drag & drop onto a window behind the current one? Start dragging, press alt-tab until it's selected, release alt, release mouse. Probably easier than the approach described in this article. Also, Windows is pretty good at providing keyboard interfaces to everything. In fact, I don't think there's anything in the core install of Windows that isn't keyboard accessible (although navigating web sites is tedious -- tab, tab, tab, tab, tab, tab, tab, tab... wait! shift-tab enter...). By comparison, most Linux desktops are positively keyboard unfriendly.
Oh, and let's not forget the fact that Windows drops internal messages occasionally anyway (about 1 in 100,000 the last time I measured it, in case anyone cares), so remember that the next time you decide to create your own message class and "trust" Windows to deliver your messages.
Interesting. How did you determine this? What messages were you using? I'm running a set of applications here that would deadlock within about 30 minutes if this was happening. They quite happily run for as long as the system is up (typically days or weeks) without problem, though. I've used this approach on NT4, W2K and XP Pro all without issue.
Slashdot Mirror
with the internet, u're confined to (a) the size of an IP packet, and (b) ur pipe.
It's no worse than telnetting to a remote server and running a keystroke interactive app (e.g. emacs). And I can confirm that that works fine, even on a 14.4k modem.
Even worse, there's no guaranteed order of delivery. It might come out "i like girls arses analiy".
a chat room (can't remember what they were called back then). I think that was 1990.
In the mid 90s we called 'em talkers. Don't know how long that term had survived for though, it might have been a relatively new one.
In my IT consultancy business, we used to use Linux desktops for a time. When we did this, we would use ytalk as a primitive IM application; it was primarily useful if one of us needed to get information from another while talking to a client on the phone. It helped present an aura of "these people know what they're doing."
We switched to Windows desktops when it became apparent that we needed to use Internet Explorer for compatibility testing on too regular a basis for any other solution, and at the time there was no obvious replacement for ytalk. We eventually started using modern IM software when it became widespread, but no longer find as much need for it.
I suppose I could further increase my speed by relying on shorthand, but I can't bring myself to butcher the language like that (eg. c u l8r!)
With predictive entry, it's actually harder to write that. On my Nokia, "c u l8r" takes 14 button presses in predictive mode, or 17 in non-predictive, whereas "see you later" takes only 13 -- plus it's much faster because getting the '8' requires going through a menu which involves waiting while the menu is displayed.
Do I need my sound card to run at kernel speed?
The hard disk driver?
Even the NIC card?
I don't think so. The CPU is spending most of its time idle on most machines, so why do drivers for SLOW HARDWARE have to be running at kernel speed?
Because that would make the hardware slower, and it is because the hardware is slower than the CPU that the CPU is spending most of its time idle (i.e. waiting for hardware to finish its job before it can get on with the next one).
Having debugged a network driver problem on one of my machines lately, I can tell you that a couple of milliseconds extra latency in responding to an IRQ from a network card can drop the maximum throughput of an SMB file server from ~50% utilisation down to ~5%. That's a huge difference. Extrapolating to a more reasonable delay of a few hundredths of a millisecond and you'll still see a noticeable performance degradation.
You'll notice it more with faster hardware. Add a hundredth of a millisecond delay to each hard disk access and you'll see file system performance drop by a significant figure, maybe as much as 10%.
Sound card -- that probably won't make much difference, except for professional recording studio applications. But they need an OS too, so we might as well keep the driver in there for them.
Latency's a killer. Doing everything you can to reduce it really increases throughput.
XP 64-bit reports itself as 5.2, because it's based on the 2K3 kernel.
Check first, post later, eh?
I figure, you'd want to replace something like WINLOGON.EXE, or whatever the closest equivalent of init there is on Windows. I'm sure there are people here who are a lot more knowledgable about how WinNT starts.
I suspect you'd have to replace either CSRSS.EXE or SMSS.EXE, and the app you replace it with would have to be a native application, so it couldn't be CMD.EXE which is a win32 console subsystem application. More info on sysinternals, here and here.
Note that I/O will be your primary difficulty -- the only API available to you was designed for output only during the blue screen phase of Windows NT's boot process, and for display BSODs. You will probably have to install a device driver that enables access to a text console and use that for IO.
This can be done, as both Windows Setup and the Recovery Console seem to use this approach.
OK -- I'm running Pro and get task manager directly from ctrl-alt-del. How do I get to the other screen you're talking about? (I used to find it useful on W2K)
You missed a few:
Wireless Zero Configuration (if you have a wireless card you should probably use the vendor's config interface -- the Windows one sends a query out for avaiable networks every minute which causes delays with some cards)
Secondary Logon, Terminal Services, Fast User Switching Compatibility (these are needed if you want to use fast user switching, but many people don't, so you can safely disable them in this case. You can disable Fast User Switching Compatibility anyway -- I've not found anything that needs it.)
Computer Browser (doesn't seem to do anything useful)
Server, Workstation (unless you need to share files or printers or use such shares respectively)
Cryptographic Services
Application Layer Gateway
Yes, lets turn off the firewall.
Brilliant.
Err... the firewall that most people use is the one in the "Windows Firewall/Internet Connection Sharing" service. I've never heard of anybody who actually does what's suggested in that link.
If you want something lighter weight than that, remove all network services other than TCP/IP from your network adapter and then enable the port-based filtering on the TCP/IP Advanced Settings dialog. You can't do source based filtering, but few people actually need to do that.
I run XP on a machine with only 96Mb of RAM. Needless to say, I've experimented a lot with which services I actually need to be running. The ones I still have are:
Automatic Updates (couldn't live without it, unfortunately)
COM+ Event System (some of the management processes use it)
DHCP Client (I could assign myself a permanent IP address, but I'm too lazy to figure out how to do that with my router)
Event Log (cause it's kind of handy sometimes)
Plug & Play
RPC (cause other stuff depends on it)
Server (I use it, you might not...)
Shell Hardware Detection
System Event Notification (I think I need this, I'm not entirely sure though)
TCP/IP NetBIOS Helper
Windows Audio
Workstation
However, the reason Firefox is slower is because it has the XPCOM-platform-abstraction-layer and uses the Javascript-bindings for core-functionality (browser.js is the actual browser; I'm not joking), which eases cross-platform development, but causes performance-penalties.
Were Opera not a cross platform browser, I'd agree that you have a point there. But, Opera have clearly shown that you can have a working browser on multiple platforms without it being slow.
My personal opinion is that too much of Firefox is implemented in JS. Probably half of that JS could be easily reimplemented as native code for a substantial speed bonus. I'd like to see that browser.
I personally have noticed no speed difference [between IE and Firefox], but I have a fast machine.
For reference, on my PII-400 I'd say firefox takes about 2-3x as long to start up, and frequently suffers long delays in various actions. Particularly grievous is the long (~200ms) pause that frequently occurs after typing the second letter of a URL in the address box while it looks up history items starting with those two letters. This pause is also noticeable on a Celeron 1.3GHz laptop, although nothing like as annoying.
Firefox also seems to use about 50% more memory on average for the same operation. It is also noticeable that it only uses single threads for many things where IE uses multiple: if one window starts a plugin, for example, all the others freeze until after the plugin has finished initialising.
Thunderbird is worst -- my entire machine grinds to a halt while it displays the new message notification window.
Even if you do notice a difference, any semi-intelligent human being knows that a 10% increase in speed isn't everything. Firefox has so much more to offer.
True, and that's why I continue to use it, despite the inconvenience. I wouldn't give up tabbed browsing for anything, for instance.
I'll be giving IE7 a try once it comes out of beta.
You're talking about the triangles in the navigation structures, right? Why do you think this is some kind of KDE innovation? I've been using images almost identical to those in web site navigation systems since some time around 1998[1]. It's hard to think of a simpler set of symbols to suggest the concepts we're talking about here. This is exactly why patents on user interfaces are so absurd -- usually any individual idea is so obvious it will be implemented hundreds of times by people who have no idea about what each other are doing.
[1] I don't know when KDE started using them like that, but I hadn't ever seen or used KDE at the time.
Yes, although it's not quite as simple as the OP said. There is a defense of not having the key: you just have to be able to provide enough evidence that you might not that it calls the prosecution's evidence that you do into doubt.
the RIP Act isn't available to enough authorities, or is too hard to invoke or something.
Err, no, not really. It's actually too easy to invoke -- a simple piece of paper is all that's required, reading something like this:
NOTE OF REQUIREMENT TO DISCLOSE ENCRYPTION KEY
You are hereby required under the Regulation of Investigatory Powers Act 2000 to provide copies of any and all encryption keys to which you have access which are assocaited with the email address suspect@example.com. You are furthermore ordered not to disclose the contents of this order to any individual except authorised representatives of the police force, and your legal adviser if any. You have until 10am tomorrow morning to comply.
Signed
(A chief police officer, a police officer authorised to sign on behalf of a chief police officer, a justice of the peace, or similar).
Ah, yes, of course that's how CSS encryption works.
BODY, TABLE {
font-family: ZapfDingbats, WingDings;
}
Good luck trying to read my documents now!
No. What happens is you get a notice and that notice tells you that you aren't allowed to reveal the existence of the notice to anyone else. Which means in your example, doing anything other than answering "no" might be a breach of the requirement.
That's not proof beyond a reasonable doubt, that's enough evidence to suggest it might be the case and it can't be proved beyond reasonable doubt the other way.
but what "400% efficiency" means in the rest of the world is "takes one unit of energy as input, and outputs four units of energy" -- i.e. free energy.
You can of course use this technique to get "free energy" (i.e. energy captured from the environment). See heating systems currently sold as "geothermal" (which aren't reallygeothermal, they just capture heat from the ground that has been warmed up by the sun).
A random thought: I thought iTunes didn't sell Beatles music, due to the trademark issues surrounding the "Apple" name related to music? Has this now been resolved?
Windows: good for getting those right-click menus. Also the only way to do things that don't have obvious keyboard shortcuts - preference dialogs, toolbar buttons, etc.
I'm a windows users who has an instinct to use the keyboard. Drag & drop onto a window behind the current one? Start dragging, press alt-tab until it's selected, release alt, release mouse. Probably easier than the approach described in this article. Also, Windows is pretty good at providing keyboard interfaces to everything. In fact, I don't think there's anything in the core install of Windows that isn't keyboard accessible (although navigating web sites is tedious -- tab, tab, tab, tab, tab, tab, tab, tab... wait! shift-tab enter...). By comparison, most Linux desktops are positively keyboard unfriendly.
Oh, and let's not forget the fact that Windows drops internal messages occasionally anyway (about 1 in 100,000 the last time I measured it, in case anyone cares), so remember that the next time you decide to create your own message class and "trust" Windows to deliver your messages.
Interesting. How did you determine this? What messages were you using? I'm running a set of applications here that would deadlock within about 30 minutes if this was happening. They quite happily run for as long as the system is up (typically days or weeks) without problem, though. I've used this approach on NT4, W2K and XP Pro all without issue.
"we have at least 60 textbook adoption software packages"
WTF is one of those, and why do you need them? Textbooks are books. You read them. Why does a computer even come into this equation?