Frankly, no. You may not force me to financially support your speech. Unless I get representation.
The USA went to WAR on this issue, for fucks sake. "No Taxation without Representation", I believe was the cry. The logical conclusion is the right to kill spammers, of course.
I much prefer Impress 3. I put the current slide, next slide, and speakers notes on the built-in screen, and the current slide on the projector. It has a built-in clock.
Big enough on the 8.9" screen for me to see.
Works better than delivering with PowerPoint, so I prefer to use it; even when delivering PowerPoint sourced presentations.
I likewise prefer Writer to Word. Spreadsheets are a bit of a wash for me. (I use Perl, Octave and other tools to reduce data to CSV form, which is then imported into a spreadsheet for final formatting; application of colour, etc). OpenOffice.org3 does VBA for more compatibility (hasn't yet been an issue), and can edit PDFs (major feature for me).
All in all, I view OpenOffice.org3 as a plus. Along with that, I use FireFox, and Evolution, so why would I bother with Windows(tm)? Linux (or any other base) that can support these applications is just fine.
Pick the one that works well and is least expensive. Would be BSD or Linux -- Windows(tm) isn't even in the running.
More for less, sounds good to me. I do save in Microsoft Office(tm) 97 formats for external presentation, which (so far) hasn't been a problem.
"What is truly sad about this is the public perception that teachers and the teacher unions are disrupting eduction. As is clearly shown here, the disruptive comes from persons who beleive they are so smart that they are not forced to be a teacher, and therefore qualified to tell the teacher what to do."
This is not clear. From either article. The disruption came from a student demonstrating Linux. The next disruptiion came from the teacher.
I also have a hard time parsing your second sentence. I will presume English is not your preferred language -- can you please restate (in your native language, or, perhaps, in simpler terms)?
And that binary format contains calls to Windows(tm) drawing routines. That WAS exploited earlier. This may be similar. Microsoft binary formats had little to do with interchange, and a lot to do with efficiency. Not having to deal with portability issues means that things like drawing dll parameters where simply "jammed in".
This increased the attack surface to a major part of the Windows(tm) drawing and rendering code.
And, since the display driver is running in kernel, it too was part of the attack surface. I had presumed that this was all straightened out, but... who knows?
The effect is that valid code sequences CAN be dropped into WORD(tm) files, and buffer overflows may extend into rather unexpected places, that can then run the exploit code. Rather clever, actually. Its a clever exploit on "clever" coding. What it should have taught is to review "clever" code, and rip it out.
On to a real-life story. A patch was developed for a performance enhancement for a BIG Unix installation. The Unix vendor (all names removed to protect the guilty and innocent) had generated the patch internally. It worked, and "saved the day" for the client (a major financial institution). The patch was proposed as an add-in for the regular Unix system; after all, there may be other clients who could benefit. The Unix committee squashed the patch -- it was too clever.
Either make the game compelling for replays, or come up with games that appeal to a wider audience.
The Wii makes some steps in this direction.
Full disclosure: I haven't played a "shooting game" since Doom. I enjoy games like "Sam and Max", but I don't see the replay value. I also enjoy Backgammon, Sudoku puzzles, and Crosswords. I get quite enough "immersive experience" with real life, thank you. So my total gaming expenditure in that last 5 years has been zero. (So why the hell am I commenting? Because I wouldn't mind seeing someone developing games that appeal to me).
Firefox 2 works fine for me, and I really see no reason to upgrade (I did have a compelling reason to move to OpenOffice.org3, though). FF2 works for me, and it is a "known quantity". FF3 isn't, and hasn't provided a single compelling reason to upgrade (yet). I upgraded from oo2 to oo3 to take advantage of: limited pdf import, read support for latest microsoft office(tm), better presentation controls (these three are very useful, and where compelling).
ff2 to ff3? I don't care about the speed of javascript, I don't have memory problems, or rendering problems. Java works ok, and I am used to the UI. I don't really care about "anti-phishing", and I find ff2's treatment of self-signed certificates annoying enough.
I sell "dual core intel computer with 2gb" for $1000 (or even more). Now, the customer *could* go to tigerdirect.ca and buy the "same" system for a few hundred dollars. I make my client VERY aware of that option. Really, I don't want any buyers remorse or anxiety over purchasing a system from me.
But... on-site setup, customized media software, lifetime labor, quality parts, little to no noise, and a nice pvr case.
Let's see tigerdirect.ca compete with that.
If *all* you are doing is selling the speakers -- I don't have much sympathy. Take your $50 dollar profit, if that's all you can get. Buy more speakers, and go "internet" as well.
MAP *does* gouge the consumer; if only to keep your business model afloat.
Personally, I think that MAP is designed to protect "reputation". Without the need for anyone to apply any extra elbow grease.
Well... I am not that familiar with the US legal system, but I really don't see how he can claim standing.
Indeed, he could be a witness (for either side) -- but if that doesn't happen, I don't see it.
So, it seems important for him to PREVENT the case from proceeding, even (or because) of his potential testimony. Indeed, what would he say? Pre-trial, "I am sorry, there are inaccuracies...", at trial "Yes, Media Sentry was engaged for investigations...".
I believe that the strategy is to supply the desired answer, while never getting into the position of being asked the "wrong" question. I further believe that the idea is to attempt to keep the issue in the "copyright infringement" domain, and out of the "unlicensed investigation" domain.
But its NOT a trick. Indeed, you just used the words "reserving memory".
Imagine a restaurant. You make a reservation for (say) 9pm. The table is yours then, but not now. The restaurant can use the same table for an early seating, or, if already occupied, doesn't have to evict the current diners.
Rplace table/diners with physical memory. Replace reservation with virtual address space.
If I understand your original argument, you are arguing for paging to disk, in order to maximize RAM usage. I believe you even argued that aggressive paging was good. I am indicating that the time required for paging (i/o) is significant, and can slow user processing for little to no gain. Effective paging is not paging, unless it is really needed. The goal for an effective paging routine is to exploit ideas like the above to make paging less painful if it is needed.
Why load code and not use it? It costs [disk time] to load code, not [memory time]. If code IS used, will it not be used again? Look at Vista -- trying to predict WHEN a program will be loaded, in an effort to amortize the time (the same idea as the paging above, but sort of in reverse). Same idea with other preload schemes.
Look at the Unix "sticky" attribute.
As to paging to disk -- a big idea is that code can be "paged out" WITHOUT writing. After all, the code pages are in the original executable file. Saves the writing! That is "good paging". The actual physical memory doesn't need to be touched UNTIL the application actually writes to it. Which means that physical memory can be both "held in reserve" and "in use for a purpose" AT THE SAME TIME (it is code RIGHT NOW, but if the other application issues a WRITE, it becomes data after an exception -- note that no disk i/o occurs at all).
I was just pointing out that a "(exclusively) keep x amount of memory for potential use" scheme is very naive, and can even be considered wrong. A good pager will accommodate these ideas (and more). Consider: EVEN if the policy were to keep an amount of memory "in reserve", why couldn't that memory keep programs that have been recently used? Better, programs that (hopefully) WILL be used? (following the model of Vista). The most effective use of memory would be to 0 (zero) bytes NOT in use at any time, and have memory to give to applications on an on-demand basis. THAT is what we are aiming for.
But -- it does make it difficult to answer the question "how much memory do I have?". It is also true that this question doesn't actually mean much, because there is a much simpler answer to the question "How efficient is my pager?". Just pick a load, and run it with paging and without, and characterize it (time spent, and amount of i/o).
If an application requests 3GB of memory (your example), it is expecting to do something to that memory. Fortunately (for performance reasons), its use of that memory is still in the future.
Basically, the request can be treated as a request for an address space and a potential allocation of memory.
The actual memory doesn't have to be supplied until the application does some with it (or, hopefully, just before). This effect can be usefully exploited to amortize the cost of paging out.
In other words, the OS doesn't have to swap out on such an allocation.
Of course, the actual availability of memory has to be accounted for. Linux uses this approach, with poor availability management:
Quoting "man malloc"
" By default, Linux follows an optimistic memory allocation strategy.
This means that when malloc() returns non-NULL there is no guarantee
that the memory really is available. This is a really bad bug. In
case it turns out that the system is out of memory, one or more pro-
cesses will be killed by the infamous OOM killer. In case Linux is
employed under circumstances where it would be less desirable to sud-
denly lose some randomly picked processes, and moreover the kernel ver-
sion is sufficiently recent, one can switch off this overcommitting
behavior using a command like:
Is the classic model. It works. I have my private network, and it provides SOME services, including email exchange.
But, the security of my "community" is my own concern. After all -- who should I trust? The government? My upstream provider? No thanks.
Now, I (as many other operators) do share information on security. But that is a voluntary activity. The buck still stops here.
The current set of laws support this attitude as well. There is a considerable base of law around "unauthorized computer access", but really nothing around "unauthorized network access". When the net was young, we all strove to deliver packets as best as we were able. Now, we are pickier about packets flowing through our systems (blocking outgoing port 25 for example, or not allowing mail relaying).
I am against a central authority. I don't see how it would work (be authoritative) and still give me the flexibility to control my own network. There are only two "commons" services that I consider mandatory; DNS and email. Email is already under attack, and we have survived (although I would like to hold in reserve the right to beat spammers to a pulp). This leaves DNS (specifically the root servers) as the most important and vulnerable common resource (from my perspective). I (my community) *could* get by with hosts lists in a pinch, though, which mitigates the risk somewhat (considering web: google/slashdot/etc. access not immediately critical to my network use).
Probably the ones I use -- tripwire (or other IDS software), SELinux (or other MAC software), or, on older systems, anti stack-smash software, like libsafe, and anti-hack software like denyhosts.
All of which are not, strictly speaking, "anti-virus" software; they simply look for, alert, or stop, misbehavior.
The reason is simple -- a government that is elected sets up an organization. Typically, the organization is created without an "exit strategy". After which, future governments end up feeding it anyway.
So, we end up with a "internet monitoring" or "media monitoring" organization. It may live on a LONG time.
As an example: Ontario, Canada, created a censorship tribunal in 1911. By the 80's, very few people could tell much about it, although it was still active. Indeed, it existed until 2004, when it was declared "unconstitional" (in the Canadian sense). See: http://www.ccla.org/news/winter04-05_10.html
However, the Ontario Film Review Board still exists (http://www.ofrb.gov.on.ca/english/page4.h) but I find it interesting that the events of 2004 are not mentioned in its "self-history".
That is what happens to these initiatives. Leading to the only solution possible. "I say we take off and nuke the entire site from orbit. It's the only way to be sure." (Aliens).
Um... a new language, a new OS (for classic HPC). The developers who ARE familiar with Windows are not (necessarily) familiar with HPC. The new language flattens everything.
I guess it's good for Microsoft -- they can try to build a "platform" or "ecosystem" out of this. But, I still don't get it. Every one of those HPC systems is different, so why would you even TRY to run a COTS OS?
Let's look at a (past) representative "bitch session":
Random Windows Luser: "Well, I heard about your OS, and it's a piece of SHIT! It won't work with my modem! Windows (95, 98) works with my modem! Any OS that won't work with my modem is SHIT!"
Reply: "But, you have a WinModem... it only works with Windows"
Luser: " is horrible! It just doesn't work"
You can replace WinModem with , or , etc.
I have had these exchanges for years; they just don't get any more pleasant. So, I have a new mantra. It's not "Microsoft is a convicted monopolist", "Windows XP is the Swiss Cheese of operating systems", (even though these are true statements).
It is simply this:
Linux (Solaris, whatever platform)... when you know WHY you want it, it will be there for you. Until then, you paid for Windows -- use it, and leave me alone. If you want a computer for games, use Windows; leave me alone. If you need Photoshop, use Windows; leave me alone.
Please don't try to engage me in OS design discussions, or implementation quality discussions. I can't help you with your Windows(tm) problems, and I am unwilling to help you with issues relating to closed source software. Take them up with your vendor.
So, yes, I now most people "Windows(tm) may be for you. Please look at Apple offerings as well".
Back to Intel: the 945 appears to be a reasonable display controller; it works fine on my Acer Aspire One. Which runs Linux -- it appears to have enough OpenGL guts to do the (admittedly limited) 3D I need for data analysis (I don't know if does Z cued lines, or Phong shading, even - but for my uses it is just fine). I don't know Vista and have no intention of ever using it, but I agree that if a computer had a "Vista Ready" sticker on, and Vista does not work properly, the customer should have recourse. If *I* purchased a system that advertised "modem" and it was (only) a WinModem, I would be upset -- but I would know within hours. It's just that I have been dragged into such discussions (hey! you're a Unix/Linux guy. Make this work! No? Well, ___ is worthless, then).
GPGPU is USELESS unless it is in hardware. It is were a "driver thing", it couldn't possibly work. Indeed, the GPGPU user would get worse performance, because the multiple execution units would have to be simulated...
So, these "optimizations" are not in the driver.
Re:WTF? If AMD64 can't do it with a full x86 core.
on
Ubuntu Ports To ARM
·
· Score: 1
Generally, just stick in winelib. Of course, the code will be ugly -- mov eax,0 would become "eax = 0;" mov ax,0 would be something like "eax &= 0xffff0000;" or "eax = (eax & 0xffff0000) | 0;" -- assuming of course C as a target. Flags are the "trickiest" to deal with. A rolling queue of generated instructions that goes back to the last instruction that could possibly have set a flag, along with the possibility of "slow" generation vs. slightly faster generation (without the flag effects).
Just some ideas... Also, look at QEMU; it does it dynamically. Statically is preferred, for performance.
If you do it... have fun with it!
Re:WTF? If AMD64 can't do it with a full x86 core.
on
Ubuntu Ports To ARM
·
· Score: 1
What problems with 64 bit? 64 bit OS can run 32 bit applications. Yes, I do use "Flash", but I haven't had the urge to run it as a 64 bit application (why would I want to?)
"Windows only stuff" -- but, we have QEMU. Now, I must confess, I use QEMU to run ARM on x86, and x86 on SPARC, but you can use it to run x86 on ARM. Now, I could be mistaken (not needing Windows(tm)), but I thought QEMU x86 translation was good enough to boot Window(tm) XP.
If QEMU isn't suitable (performance), I would assume that another binary translator will be produced. (its just that QEMU is the -arguably- most popular binary translator in use). For performance, the WINE libraries could be deployed to make the "under-the-cover" APIs work at full native speed.
Actually, I would think that most Windows(tm) software does not use self-modifying code (for security reasons), and, as a result, could be statically recompiled, and re-linked against alternate (WINE-like) libraries. The technique is currently being deployed on old MS software (see/. for the story on deploying CBM BASIC, which is a statically recompiled 6502 BASIC). I probably wouldn't use "C" as my intermediate language for this (SCHEME strikes me as a better choice), but it is well within the capabilities of "amateur" developers to do this.
The largest impediment is the copyright of the original work. It may take quite a while to recompile such an application, and the results would be a "derivative work". In other words, not legal to distribute. The recompiler itself would be legal. You probably won't be sued for distributing a recompilation of CBM BASIC, but MS Word of recent vintage?
\\.\PhysicalDriveN for physical drives (0, 1, 2...), or \\.\X: for logical drives (C, D, E...). Of course, the mapping between physical and logical isn't particularly clear, given the partition tables on the physical devices.
Under Unix (linux, here)/dev/sdx is physical drive x (a b c...),/dev/sdxn (1 2 3...) is a partition. Its indeed a bit easier (insertion of an "earlier" physical drive doesn't change the designations of other drives).
But, even so, as "Administrator" on Windows, if you know the physical or logical drive, you can use the "COPY" command to duplicate the drive (I think -- never dared).
COPY \\.\E: C:\IMAGE REM THERE IS NO STANDARD COMPRESSOR UNDER WINDOWS, USE ZIP
"TYPE" may work to stream the data, oh, and you still have to replace \\.\E: with the designator for the partition (whatever it is).
Make sure the disk is zeroed prior to installing anything (dd if=/dev/zero of=/dev/sdb -- replacing sdb with whatever the drive actually is).
The partition the drive and install your software.
To capture -- dd if=/dev/sdb | bzip2 >image.bz2
I would use bzip2 instead of gzip for the slightly better compression. It would be possible to "delta" two images, but you didn't ask about it.
This presumes unix (linux), possibly as a "live cd"; it may be workable with "cygwin".
It is ALSO possible to capture just a single partition. Indeed, I would recommend this approach. Assuming you are using the first partition, partition the drive first. dd if=/dev/zero of=/dev/sdb1 to zero the drive. Install into that partition, and then dd if=dev/sdb1 | bzip2 >image.bz2 to capture.
MAKE SURE THE RECEIVING FILE SYSTEM IS NOT FAT32! You need something that will allow large files. ntfs or ext2/3 or something like that will work.
The image.bz2 file can be decompressed and used with (say) VMware directly, or it can be mounted (mount -t ntfs -o loop image/mnt would work under linux).
I am sure that we will get plenty of "defensive" posts from Windows users.
So let's just nip before it buds. Firstly, driver support in Linux is supported by the systems that are supported by the OS. For Linux, the number of systems is somewhere over 80.
Of course, the argument will then be "But the only useful system to me is x86, perhaps x86_64". How many drivers are included? For Linux 2.4, Redhat supplied over 700 drivers (not counting filesystems, and kernel modules as drivers). By Fedora 8, over 1100 drivers are supplied in a default configuration. Note that these are x86 platform only.
There are parts of the driver matrix which are just not sensible. An example would be Apple ADB on Intel. The 700/1100 counts above do NOT include those "silly" possibilities -- these are physically shipped drivers (cd/lib/modules/kernel*/drivers; file . -name "*.ko" | wc).
I am interested -- how many drivers are supplied with Vista?
So you think the convention broke the law? Why? Well... they collected money....
I see. Vendor permit notwithstanding, I imagine it would come down to tax receipts, right? Spin, spin, spin.
Because that isn't the point. The point is that the CCF should have provided an understandable and reasonable explanation. The consequence to the CCF for NOT doing that? I call them dunderheads.
For the CCF to reverse sponsorship roles in their explanation? Looked like "instant-spin" to me!
What do they take ME for? Stupid, I guess. Well, I *used* to be a CCF supporter...
Slashdot Mirror
Frankly, no. You may not force me to financially support your speech. Unless I get representation.
The USA went to WAR on this issue, for fucks sake. "No Taxation without Representation", I believe was the cry. The logical conclusion is the right to kill spammers, of course.
I much prefer Impress 3. I put the current slide, next slide, and speakers notes on the built-in screen, and the current slide on the projector. It has a built-in clock.
Big enough on the 8.9" screen for me to see.
Works better than delivering with PowerPoint, so I prefer to use it; even when delivering PowerPoint sourced presentations.
I likewise prefer Writer to Word. Spreadsheets are a bit of a wash for me. (I use Perl, Octave and other tools to reduce data to CSV form, which is then imported into a spreadsheet for final formatting; application of colour, etc). OpenOffice.org3 does VBA for more compatibility (hasn't yet been an issue), and can edit PDFs (major feature for me).
All in all, I view OpenOffice.org3 as a plus. Along with that, I use FireFox, and Evolution, so why would I bother with Windows(tm)? Linux (or any other base) that can support these applications is just fine.
Pick the one that works well and is least expensive. Would be BSD or Linux -- Windows(tm) isn't even in the running.
More for less, sounds good to me. I do save in Microsoft Office(tm) 97 formats for external presentation, which (so far) hasn't been a problem.
So, "Linux is usable for Business" for me.
"What is truly sad about this is the public perception that teachers and the teacher unions are disrupting eduction. As is clearly shown here, the disruptive comes from persons who beleive they are so smart that they are not forced to be a teacher, and therefore qualified to tell the teacher what to do."
This is not clear. From either article. The disruption came from a student demonstrating Linux. The next disruptiion came from the teacher.
I also have a hard time parsing your second sentence. I will presume English is not your preferred language -- can you please restate (in your native language, or, perhaps, in simpler terms)?
And that binary format contains calls to Windows(tm) drawing routines. That WAS exploited earlier. This may be similar. Microsoft binary formats had little to do with interchange, and a lot to do with efficiency. Not having to deal with portability issues means that things like drawing dll parameters where simply "jammed in".
This increased the attack surface to a major part of the Windows(tm) drawing and rendering code.
And, since the display driver is running in kernel, it too was part of the attack surface. I had presumed that this was all straightened out, but... who knows?
The effect is that valid code sequences CAN be dropped into WORD(tm) files, and buffer overflows may extend into rather unexpected places, that can then run the exploit code. Rather clever, actually. Its a clever exploit on "clever" coding. What it should have taught is to review "clever" code, and rip it out.
On to a real-life story. A patch was developed for a performance enhancement for a BIG Unix installation. The Unix vendor (all names removed to protect the guilty and innocent) had generated the patch internally. It worked, and "saved the day" for the client (a major financial institution). The patch was proposed as an add-in for the regular Unix system; after all, there may be other clients who could benefit. The Unix committee squashed the patch -- it was too clever.
And that's what SHOULD happen.
The solution is not preventing used games sales.
Either make the game compelling for replays, or come up with games that appeal to a wider audience.
The Wii makes some steps in this direction.
Full disclosure: I haven't played a "shooting game" since Doom. I enjoy games like "Sam and Max", but I don't see the replay value. I also enjoy Backgammon, Sudoku puzzles, and Crosswords. I get quite enough "immersive experience" with real life, thank you. So my total gaming expenditure in that last 5 years has been zero. (So why the hell am I commenting? Because I wouldn't mind seeing someone developing games that appeal to me).
Faster? Has the fsync issue been resolved yet?
Firefox 2 works fine for me, and I really see no reason to upgrade (I did have a compelling reason to move to OpenOffice.org3, though). FF2 works for me, and it is a "known quantity". FF3 isn't, and hasn't provided a single compelling reason to upgrade (yet). I upgraded from oo2 to oo3 to take advantage of: limited pdf import, read support for latest microsoft office(tm), better presentation controls (these three are very useful, and where compelling).
ff2 to ff3? I don't care about the speed of javascript, I don't have memory problems, or rendering problems. Java works ok, and I am used to the UI. I don't really care about "anti-phishing", and I find ff2's treatment of self-signed certificates annoying enough.
Is there something I am missing?
I sell "dual core intel computer with 2gb" for $1000 (or even more). Now, the customer *could* go to tigerdirect.ca and buy the "same" system for a few hundred dollars. I make my client VERY aware of that option. Really, I don't want any buyers remorse or anxiety over purchasing a system from me.
But... on-site setup, customized media software, lifetime labor, quality parts, little to no noise, and a nice pvr case.
Let's see tigerdirect.ca compete with that.
If *all* you are doing is selling the speakers -- I don't have much sympathy. Take your $50 dollar profit, if that's all you can get. Buy more speakers, and go "internet" as well.
MAP *does* gouge the consumer; if only to keep your business model afloat.
Personally, I think that MAP is designed to protect "reputation". Without the need for anyone to apply any extra elbow grease.
Well... I am not that familiar with the US legal system, but I really don't see how he can claim standing.
Indeed, he could be a witness (for either side) -- but if that doesn't happen, I don't see it.
So, it seems important for him to PREVENT the case from proceeding, even (or because) of his potential testimony. Indeed, what would he say? Pre-trial, "I am sorry, there are inaccuracies...", at trial "Yes, Media Sentry was engaged for investigations...".
I believe that the strategy is to supply the desired answer, while never getting into the position of being asked the "wrong" question. I further believe that the idea is to attempt to keep the issue in the "copyright infringement" domain, and out of the "unlicensed investigation" domain.
Monads are incredibly inefficient?
But its NOT a trick. Indeed, you just used the words "reserving memory".
Imagine a restaurant. You make a reservation for (say) 9pm. The table is yours then, but not now. The restaurant can use the same table for an early seating, or, if already occupied, doesn't have to evict the current diners.
Rplace table/diners with physical memory. Replace reservation with virtual address space.
If I understand your original argument, you are arguing for paging to disk, in order to maximize RAM usage. I believe you even argued that aggressive paging was good. I am indicating that the time required for paging (i/o) is significant, and can slow user processing for little to no gain. Effective paging is not paging, unless it is really needed. The goal for an effective paging routine is to exploit ideas like the above to make paging less painful if it is needed.
Why load code and not use it? It costs [disk time] to load code, not [memory time]. If code IS used, will it not be used again? Look at Vista -- trying to predict WHEN a program will be loaded, in an effort to amortize the time (the same idea as the paging above, but sort of in reverse). Same idea with other preload schemes.
Look at the Unix "sticky" attribute.
As to paging to disk -- a big idea is that code can be "paged out" WITHOUT writing. After all, the code pages are in the original executable file. Saves the writing! That is "good paging". The actual physical memory doesn't need to be touched UNTIL the application actually writes to it. Which means that physical memory can be both "held in reserve" and "in use for a purpose" AT THE SAME TIME (it is code RIGHT NOW, but if the other application issues a WRITE, it becomes data after an exception -- note that no disk i/o occurs at all).
I was just pointing out that a "(exclusively) keep x amount of memory for potential use" scheme is very naive, and can even be considered wrong. A good pager will accommodate these ideas (and more). Consider: EVEN if the policy were to keep an amount of memory "in reserve", why couldn't that memory keep programs that have been recently used? Better, programs that (hopefully) WILL be used? (following the model of Vista). The most effective use of memory would be to 0 (zero) bytes NOT in use at any time, and have memory to give to applications on an on-demand basis. THAT is what we are aiming for.
But -- it does make it difficult to answer the question "how much memory do I have?". It is also true that this question doesn't actually mean much, because there is a much simpler answer to the question "How efficient is my pager?". Just pick a load, and run it with paging and without, and characterize it (time spent, and amount of i/o).
If an application requests 3GB of memory (your example), it is expecting to do something to that memory. Fortunately (for performance reasons), its use of that memory is still in the future.
Basically, the request can be treated as a request for an address space and a potential allocation of memory.
The actual memory doesn't have to be supplied until the application does some with it (or, hopefully, just before). This effect can be usefully exploited to amortize the cost of paging out.
In other words, the OS doesn't have to swap out on such an allocation.
Of course, the actual availability of memory has to be accounted for. Linux uses this approach, with poor availability management:
Quoting "man malloc"
" By default, Linux follows an optimistic memory allocation strategy.
This means that when malloc() returns non-NULL there is no guarantee
that the memory really is available. This is a really bad bug. In
case it turns out that the system is out of memory, one or more pro-
cesses will be killed by the infamous OOM killer. In case Linux is
employed under circumstances where it would be less desirable to sud-
denly lose some randomly picked processes, and moreover the kernel ver-
sion is sufficiently recent, one can switch off this overcommitting
behavior using a command like:
# echo 2 > /proc/sys/vm/overcommit_memory
"
Is the classic model. It works. I have my private network, and it provides SOME services, including email exchange.
But, the security of my "community" is my own concern. After all -- who should I trust? The government? My upstream provider? No thanks.
Now, I (as many other operators) do share information on security. But that is a voluntary activity. The buck still stops here.
The current set of laws support this attitude as well. There is a considerable base of law around "unauthorized computer access", but really nothing around "unauthorized network access". When the net was young, we all strove to deliver packets as best as we were able. Now, we are pickier about packets flowing through our systems (blocking outgoing port 25 for example, or not allowing mail relaying).
I am against a central authority. I don't see how it would work (be authoritative) and still give me the flexibility to control my own network. There are only two "commons" services that I consider mandatory; DNS and email. Email is already under attack, and we have survived (although I would like to hold in reserve the right to beat spammers to a pulp). This leaves DNS (specifically the root servers) as the most important and vulnerable common resource (from my perspective). I (my community) *could* get by with hosts lists in a pinch, though, which mitigates the risk somewhat (considering web: google/slashdot/etc. access not immediately critical to my network use).
Probably the ones I use -- tripwire (or other IDS software), SELinux (or other MAC software), or, on older systems, anti stack-smash software, like libsafe, and anti-hack software like denyhosts.
All of which are not, strictly speaking, "anti-virus" software; they simply look for, alert, or stop, misbehavior.
No, not in the "voting booth".
The reason is simple -- a government that is elected sets up an organization. Typically, the organization is created without an "exit strategy". After which, future governments end up feeding it anyway.
So, we end up with a "internet monitoring" or "media monitoring" organization. It may live on a LONG time.
As an example: Ontario, Canada, created a censorship tribunal in 1911. By the 80's, very few people could tell much about it, although it was still active. Indeed, it existed until 2004, when it was declared "unconstitional" (in the Canadian sense). See: http://www.ccla.org/news/winter04-05_10.html
However, the Ontario Film Review Board still exists (http://www.ofrb.gov.on.ca/english/page4.h) but I find it interesting that the events of 2004 are not mentioned in its "self-history".
That is what happens to these initiatives. Leading to the only solution possible. "I say we take off and nuke the entire site from orbit. It's the only way to be sure." (Aliens).
Um... a new language, a new OS (for classic HPC). The developers who ARE familiar with Windows are not (necessarily) familiar with HPC. The new language flattens everything.
I guess it's good for Microsoft -- they can try to build a "platform" or "ecosystem" out of this. But, I still don't get it. Every one of those HPC systems is different, so why would you even TRY to run a COTS OS?
What else does Windows offer here?
Wow
Let's look at a (past) representative "bitch session":
Random Windows Luser: "Well, I heard about your OS, and it's a piece of SHIT! It won't work with my modem! Windows (95, 98) works with my modem! Any OS that won't work with my modem is SHIT!"
Reply: "But, you have a WinModem... it only works with Windows"
Luser: " is horrible! It just doesn't work"
You can replace WinModem with , or , etc.
I have had these exchanges for years; they just don't get any more pleasant. So, I have a new mantra. It's not "Microsoft is a convicted monopolist", "Windows XP is the Swiss Cheese of operating systems", (even though these are true statements).
It is simply this:
Linux (Solaris, whatever platform)... when you know WHY you want it, it will be there for you. Until then, you paid for Windows -- use it, and leave me alone. If you want a computer for games, use Windows; leave me alone. If you need Photoshop, use Windows; leave me alone.
Please don't try to engage me in OS design discussions, or implementation quality discussions. I can't help you with your Windows(tm) problems, and I am unwilling to help you with issues relating to closed source software. Take them up with your vendor.
So, yes, I now most people "Windows(tm) may be for you. Please look at Apple offerings as well".
Back to Intel: the 945 appears to be a reasonable display controller; it works fine on my Acer Aspire One. Which runs Linux -- it appears to have enough OpenGL guts to do the (admittedly limited) 3D I need for data analysis (I don't know if does Z cued lines, or Phong shading, even - but for my uses it is just fine). I don't know Vista and have no intention of ever using it, but I agree that if a computer had a "Vista Ready" sticker on, and Vista does not work properly, the customer should have recourse. If *I* purchased a system that advertised "modem" and it was (only) a WinModem, I would be upset -- but I would know within hours. It's just that I have been dragged into such discussions (hey! you're a Unix/Linux guy. Make this work! No? Well, ___ is worthless, then).
Makes me sensitive.
Think.
GPGPU is USELESS unless it is in hardware. It is were a "driver thing", it couldn't possibly work. Indeed, the GPGPU user would get worse performance, because the multiple execution units would have to be simulated...
So, these "optimizations" are not in the driver.
Generally, just stick in winelib. Of course, the code will be ugly -- mov eax,0 would become "eax = 0;" mov ax,0 would be something like "eax &= 0xffff0000;" or "eax = (eax & 0xffff0000) | 0;" -- assuming of course C as a target. Flags are the "trickiest" to deal with. A rolling queue of generated instructions that goes back to the last instruction that could possibly have set a flag, along with the possibility of "slow" generation vs. slightly faster generation (without the flag effects).
Just some ideas... Also, look at QEMU; it does it dynamically. Statically is preferred, for performance.
If you do it... have fun with it!
What problems with 64 bit? 64 bit OS can run 32 bit applications. Yes, I do use "Flash", but I haven't had the urge to run it as a 64 bit application (why would I want to?)
"Windows only stuff" -- but, we have QEMU. Now, I must confess, I use QEMU to run ARM on x86, and x86 on SPARC, but you can use it to run x86 on ARM. Now, I could be mistaken (not needing Windows(tm)), but I thought QEMU x86 translation was good enough to boot Window(tm) XP.
If QEMU isn't suitable (performance), I would assume that another binary translator will be produced. (its just that QEMU is the -arguably- most popular binary translator in use). For performance, the WINE libraries could be deployed to make the "under-the-cover" APIs work at full native speed.
Actually, I would think that most Windows(tm) software does not use self-modifying code (for security reasons), and, as a result, could be statically recompiled, and re-linked against alternate (WINE-like) libraries. The technique is currently being deployed on old MS software (see /. for the story on deploying CBM BASIC, which is a statically recompiled 6502 BASIC).
I probably wouldn't use "C" as my intermediate language for this (SCHEME strikes me as a better choice), but it is well within the capabilities of "amateur" developers to do this.
The largest impediment is the copyright of the original work. It may take quite a while to recompile such an application, and the results would be a "derivative work". In other words, not legal to distribute. The recompiler itself would be legal. You probably won't be sued for distributing a recompilation of CBM BASIC, but MS Word of recent vintage?
Easy... is relative. Let's try to remove Unix (Linux) from the equation:
There is a weird convention used with Windows for direct drive access:
http://support.microsoft.com/kb/100027
\\.\PhysicalDriveN for physical drives (0, 1, 2...), or \\.\X: for logical drives (C, D, E...). Of course, the mapping between physical and logical isn't particularly clear, given the partition tables on the physical devices.
Under Unix (linux, here) /dev/sdx is physical drive x (a b c...), /dev/sdxn (1 2 3...) is a partition. Its indeed a bit easier (insertion of an "earlier" physical drive doesn't change the designations of other drives).
But, even so, as "Administrator" on Windows, if you know the physical or logical drive, you can use the "COPY" command to duplicate the drive (I think -- never dared).
COPY \\.\E: C:\IMAGE
REM THERE IS NO STANDARD COMPRESSOR UNDER WINDOWS, USE ZIP
"TYPE" may work to stream the data, oh, and you still have to replace \\.\E: with the designator for the partition (whatever it is).
Does this help?
Sure, I'll "second" that.
Make sure the disk is zeroed prior to installing anything (dd if=/dev/zero of=/dev/sdb -- replacing sdb with whatever the drive actually is).
The partition the drive and install your software.
To capture -- dd if=/dev/sdb | bzip2 >image.bz2
I would use bzip2 instead of gzip for the slightly better compression. It would be possible to "delta" two images, but you didn't ask about it.
This presumes unix (linux), possibly as a "live cd"; it may be workable with "cygwin".
It is ALSO possible to capture just a single partition. Indeed, I would recommend this approach. Assuming you are using the first partition, partition the drive first. dd if=/dev/zero of=/dev/sdb1 to zero the drive. Install into that partition, and then dd if=dev/sdb1 | bzip2 >image.bz2 to capture.
MAKE SURE THE RECEIVING FILE SYSTEM IS NOT FAT32! You need something that will allow large files. ntfs or ext2/3 or something like that will work.
The image.bz2 file can be decompressed and used with (say) VMware directly, or it can be mounted (mount -t ntfs -o loop image /mnt would work under linux).
But those organizations do license counting -- Microsoft knows how many Vista are in use at those sites.
It's the Vista bundles we are talking about.
I am sure that we will get plenty of "defensive" posts from Windows users.
So let's just nip before it buds. Firstly, driver support in Linux is supported by the systems that are supported by the OS. For Linux, the number of systems is somewhere over 80.
Of course, the argument will then be "But the only useful system to me is x86, perhaps x86_64". How many drivers are included? For Linux 2.4, Redhat supplied over 700 drivers (not counting filesystems, and kernel modules as drivers). By Fedora 8, over 1100 drivers are supplied in a default configuration. Note that these are x86 platform only.
There are parts of the driver matrix which are just not sensible. An example would be Apple ADB on Intel. The 700/1100 counts above do NOT include those "silly" possibilities -- these are physically shipped drivers (cd /lib/modules/kernel*/drivers; file . -name "*.ko" | wc).
I am interested -- how many drivers are supplied with Vista?
So you think the convention broke the law? Why? Well... they collected money....
I see. Vendor permit notwithstanding, I imagine it would come down to tax receipts, right? Spin, spin, spin.
Because that isn't the point. The point is that the CCF should have provided an understandable and reasonable explanation. The consequence to the CCF for NOT doing that? I call them dunderheads.
For the CCF to reverse sponsorship roles in their explanation? Looked like "instant-spin" to me!
What do they take ME for? Stupid, I guess. Well, I *used* to be a CCF supporter...
Sure. They can turn down anything. Personally, I don't care. In my opinion, this particular action makes them dunderheads.
Oh, the money DID go to another charity.