That's incredibly dumb. FreeBSD, for example, uses slow times stretches to pre-emptively copy idle resident processes to SWAP - not moving them, but copying them - so that a lot of RAM is available on an instant's notice should a process suddenly need it.
On a desktop that sounds like a really useful trick. On a server where the maximum VM requirements are known to be significantly smaller than the available RAM there is never going to be a time where any process will ever need more RAM than can be immediately allocated.
On some of these systems the total "disk" space has been the same order of magnitude as RAM, *and* it's flash... you don't want to write to that when you don't need to. If you mount root "noatime" you can go *days* without writing to disk.
VISTA performa aggressive background prefetching of commonly used applications. It also builds a prefetch file in the background. This takes a big chunk of memory and a big of HD space.
OK, for the sake of argument I'll take it as given that it's going to do that. I have some issues with aggressive prefetching, but for a desktop environment it seems to be a win, so I'll set that aside. For a server, where you typically don't launch large applications for short periods of time, that's a different matter.
I use FreeBSD, which has had a unified buffer cache for years. It uses all available memory as buffers, automatically. You can always drop infrequently used *clean* pages because they can get paged back in from the executable (Vista would use the prefetch cache). Dropping stale clean pages is always more efficient than paging out stale dirty pages.
Pages that are involved in your prefetch cache and other background optimizations are not candidates for paging out to the swap file. If they are, then you've got them paged out multiple places, and that's inefficient. The prefatch cache itself is the backing store for them.
So the only pages that are candidates for paging out are stale dirty pages. If I run without a swap file, then my stale dirty pages won't get paged out. That might reduce the efficiency of the background optimization, but that's not grounds for requiring the pagefile to be there at all. AND it doesn't apply to earlier versions of NT, which ALSO had this requirement.
So while that might be a good argument for having a pagefile in Vista, that's not why NT requires a pagefile, period. And that's what I'm asking... why does NT do that? What are the technical reasons that NT is not just "less efficient" but downright unstable when it doesn't have a pagefile?
However, if your disk starts getting over 80% full or you get more than 20% fragmentation on the HD, then your performance begins to suffer due to seeking.
Right, now why is that? UFS only demands 10% overhead, and I've run it with the overhead set as low as 3% for long periods of time, even on partitions that have a super-high rate of creation of small files... like news spools... without fragmentation ever getting over 1%. There are other file system designs where fragmentation is a non-issue. The only major modern operating system where fragmentation is even something you'll EVER think about is NTFS. Why is that?
I'm not asking about why Vista might be better than XP or whatever, Im askin, thechnically, what are the design elements of NTFS that make fragmentation an issue.
Seriously, I'm glad to see this. Matt Dillon is a brilliant programmer with much broader OS experience than most of the folks driving OS development these days, so I have no doubt that he's doing amazing things.
That's kind of a metaphor for life, you know. There's no way out of the canyon for any of us, you're going to hit the wall anyway, you might as well whoop and holler and get a good last line on the cockpit voice recorder.
I don't think this is comparable to the theoretical 10GB RAM disk.
No shit, Sherlock, it's the cheapest hardware RAM disk on the market, and it's three years old, of course it's small. It IS big enough to install Windows 2000 or XP on, even if Vista is too bloated, but that's not what I use it for any more. You know what I put on it? The bloody page file. Why does Windows demand a page file as big as my main memory, when even running 3d games and Photoshop at the same time never brings my working set anywhere NEAR the amount of RAM I have? I don't know. Why does a Vista install need more disk space than our main development server had in 1992? I don't know. I don't know what the hell is in there. Sentient programs looking for Megadeth rips and pirated copies of Office, for all I know. It's a bloody mystery.
But enough ranting, back to the point: you *can* get hardware RAM disks that are big enough to install the OS on.
Unless your working set is larger than your physical RAM, you shouldn't need a pagefile or swap partition at all. I routinely run UNIX based servers with no swap partition, because I know the maximum working set is much smaller than the available RAM.
I remember being puzzled by this on early versions of NT, but I figured Microsoft would eventually get around to fixing this obvious bug. Can someone explain exactly what their logic is?
Oh, and why does NTFS need to be defragmented? Building a file system that avoids fragmentation problems is not rocket science.
According to what I've googled, this is based on what was originally called "RedOffice", the Chinese fork of OpenOffice.
If UOF is based on a product derived from "RedOffice", that means the format is likely to have similar limitations to ODF and OOXML, both of which are based directly or indirectly on Word's document structure.
The problem is that Word's document structure is awful. It's not a hierarchical format in any meaningful sense, the only nestable structure is the table, and the basic block is a fully qualified paragraph or a fully qualified style. Every Word-compatible word processor I've used has had this same problem, and it makes any kind of automated processing of the resulting document a pain in the neck... or forced blind reliance on complex and opaque libraries.
Even raw HTML is a better format for documents than anything based on Word's document structure. Heck, even for archival purposes where layout is critical, Postscript or simple PDF is a better choice than a Word-derived format: it retains details of the layout and you don't need an editable format for archiving, and it's well defined with multiple existing implementations.
What the world needs is a good word processor that uses something like Docbook as its native format.
I'm saying you don't have a 10GB RAM disk, and your assertion that Vista makes RAM run poorly is pure conjecture. Let's say you somehow made a 10GB RAM disk, ignoring for the moment that you would have no RAM to run system, why you would go about installing any programs on a RAM disk is really amazingly stupid, the moment you restarted your computer the program would be annaliated.
But you don't even need a battery backed SATA RAM drive to make a RAM disk that survives reboots. The Amiga Recoverable RAM Disk (RRD.DEVICE) hooked into the hardware boot vectors and did an integrity check at start, and so long as you hadn't lost power it would re-install itself with all its contents intact. You could even use it as your boot disk.
Oh, man, if that holds up as a precedent... the implications are breathtaking.
I just felt a disturbance in the force, as if every patent troll in the country suddenly started hyperventilating and reaching for their blood pressure medication.
Then might I recommend "The Book of the New Sun", starting with "The Shadow of the Torturer". See if you can count the number of words Severian uses to describe his profession...
What does "[c]laims which are broad enough to read on obvious subject matter are unpatentable even though they also read on nonobvious subject matter." mean? It almost makes sense, but the term "read on" appears to be legal jargon, because it breaks/brain/lib/english_parser.so for me.
No, reading the Baroque cycle is like strapping yourself to a tree [...]
I didn't strap myself to that tree (or at least I cut myself loose before the end of the first volume). I'm talking about his science fiction, not his Swiss Family Robinson version of the 17th century.
Why not skip the idea of a separate browser "application" altogether and build web-rendering into something that resembles the "desktop,"
Microsoft did that in 1997. It was called "Active Desktop" and led to the biggest malware ecosystem ever.
The iPhone does not make that mistake. The browser is a separate application and the application, not the rendering engine, manages access between the sandbox and the outside world.
The iPhone costs three times as much as this device. If you subsidize it with a phone contract the list price is comparable, but then you have to add well over a thousand dollars per year for the contract.
Microsoft's problem in handhelds is Windows. They don't want the Windows Mobile based devices to become laptop replacements, because that would compete with Windows sales, but they want them to be recognizably Windows to both make development easier and to promote the brand.
Windows Mobile loses because Windows CE is just not reliable and solid enough to serve both the needs of a mobile phone and the needs of a general-purpose handheld. Palm didn't have this problem nearly as badly because PalmOS ran under a real time OS (AMX) that you couldn't get into from user applications... the whole Palm environment is just one task for AMX.
Take away the phone, and just worry about making a PDA, and you get a lot more freedom. The iPod Touch has really got more potential to benefit from iPhone apps than the iPhone, because it's not such a critical device. In the Pocket PC WinCE would have been fine if ActiveSync worked as well as Palm HotSync, so your ActiveSync repository served as a complete backup for everything on your Pocket PC... losing data every time I ran my battery flat was what drove me back to PalmOS for my PDA. So the Pocket PC loses because Microsoft didn't make it good enough to run standalone, and didn't make PC-side software good enough that you didn't care.
Going to a tablet, and you get even more freedom. Before "Tablet PC" there were Windows CE based clamshells and tablets that were quite capable, but Microsoft pretty much nuked them by loading the Pocket PC software down with restrictions (both technical and contractual) that meant the Windows CE based tablets were stuck with the previous generation of Windows Mobile software. Of course, they wanted their flagship product on the Tablet PC, not this stripped down embedded-only Windows CE.
I don't know if a browser-only tablet is a good solution, but a tablet is so far from the iPhone or Windows Mobile that trying to draw analogies between them is misleading at best... even if Microsoft hadn't continually undercut Windows Mobile to keep it from even potentially cannibalizing their flagship product.
Reading a Neal Stephenson novel is like strapping yourself into the back seat of a converted jet trainer to tour the Grand Canyon. For a lot of people, by the time they've gotten used to dodging pillars of rock at half the speed of sound and they're really enjoying the view the pilot flips over the rim and... that's all, tour's over.
I get used to the view pretty quick, and I've come to accept the endings, so I'll be picking up ANATHEM anyway.
Slashdot Mirror
Walnut Creek CDROM ran for years selling CDs that you could download from them for free.
That's how many people donated to Slackware and FreeBSD, by buying the CDROM from Walnut Creek instead of downloading the ISOs from Walnut Creek.
Apparently, Vector is just taking this a step further.
That's incredibly dumb. FreeBSD, for example, uses slow times stretches to pre-emptively copy idle resident processes to SWAP - not moving them, but copying them - so that a lot of RAM is available on an instant's notice should a process suddenly need it.
On a desktop that sounds like a really useful trick. On a server where the maximum VM requirements are known to be significantly smaller than the available RAM there is never going to be a time where any process will ever need more RAM than can be immediately allocated.
On some of these systems the total "disk" space has been the same order of magnitude as RAM, *and* it's flash... you don't want to write to that when you don't need to. If you mount root "noatime" you can go *days* without writing to disk.
VISTA performa aggressive background prefetching of commonly used applications. It also builds a prefetch file in the background. This takes a big chunk of memory and a big of HD space.
OK, for the sake of argument I'll take it as given that it's going to do that. I have some issues with aggressive prefetching, but for a desktop environment it seems to be a win, so I'll set that aside. For a server, where you typically don't launch large applications for short periods of time, that's a different matter.
I use FreeBSD, which has had a unified buffer cache for years. It uses all available memory as buffers, automatically. You can always drop infrequently used *clean* pages because they can get paged back in from the executable (Vista would use the prefetch cache). Dropping stale clean pages is always more efficient than paging out stale dirty pages.
Pages that are involved in your prefetch cache and other background optimizations are not candidates for paging out to the swap file. If they are, then you've got them paged out multiple places, and that's inefficient. The prefatch cache itself is the backing store for them.
So the only pages that are candidates for paging out are stale dirty pages. If I run without a swap file, then my stale dirty pages won't get paged out. That might reduce the efficiency of the background optimization, but that's not grounds for requiring the pagefile to be there at all. AND it doesn't apply to earlier versions of NT, which ALSO had this requirement.
So while that might be a good argument for having a pagefile in Vista, that's not why NT requires a pagefile, period. And that's what I'm asking... why does NT do that? What are the technical reasons that NT is not just "less efficient" but downright unstable when it doesn't have a pagefile?
However, if your disk starts getting over 80% full or you get more than 20% fragmentation on the HD, then your performance begins to suffer due to seeking.
Right, now why is that? UFS only demands 10% overhead, and I've run it with the overhead set as low as 3% for long periods of time, even on partitions that have a super-high rate of creation of small files... like news spools... without fragmentation ever getting over 1%. There are other file system designs where fragmentation is a non-issue. The only major modern operating system where fragmentation is even something you'll EVER think about is NTFS. Why is that?
I'm not asking about why Vista might be better than XP or whatever, Im askin, thechnically, what are the design elements of NTFS that make fragmentation an issue.
But will it run AmigaOS?
Seriously, I'm glad to see this. Matt Dillon is a brilliant programmer with much broader OS experience than most of the folks driving OS development these days, so I have no doubt that he's doing amazing things.
That's kind of a metaphor for life, you know. There's no way out of the canyon for any of us, you're going to hit the wall anyway, you might as well whoop and holler and get a good last line on the cockpit voice recorder.
I don't think this is comparable to the theoretical 10GB RAM disk.
No shit, Sherlock, it's the cheapest hardware RAM disk on the market, and it's three years old, of course it's small. It IS big enough to install Windows 2000 or XP on, even if Vista is too bloated, but that's not what I use it for any more. You know what I put on it? The bloody page file. Why does Windows demand a page file as big as my main memory, when even running 3d games and Photoshop at the same time never brings my working set anywhere NEAR the amount of RAM I have? I don't know. Why does a Vista install need more disk space than our main development server had in 1992? I don't know. I don't know what the hell is in there. Sentient programs looking for Megadeth rips and pirated copies of Office, for all I know. It's a bloody mystery.
But enough ranting, back to the point: you *can* get hardware RAM disks that are big enough to install the OS on.
Unless your working set is larger than your physical RAM, you shouldn't need a pagefile or swap partition at all. I routinely run UNIX based servers with no swap partition, because I know the maximum working set is much smaller than the available RAM.
I remember being puzzled by this on early versions of NT, but I figured Microsoft would eventually get around to fixing this obvious bug. Can someone explain exactly what their logic is?
Oh, and why does NTFS need to be defragmented? Building a file system that avoids fragmentation problems is not rocket science.
According to what I've googled, this is based on what was originally called "RedOffice", the Chinese fork of OpenOffice.
If UOF is based on a product derived from "RedOffice", that means the format is likely to have similar limitations to ODF and OOXML, both of which are based directly or indirectly on Word's document structure.
The problem is that Word's document structure is awful. It's not a hierarchical format in any meaningful sense, the only nestable structure is the table, and the basic block is a fully qualified paragraph or a fully qualified style. Every Word-compatible word processor I've used has had this same problem, and it makes any kind of automated processing of the resulting document a pain in the neck... or forced blind reliance on complex and opaque libraries.
Even raw HTML is a better format for documents than anything based on Word's document structure. Heck, even for archival purposes where layout is critical, Postscript or simple PDF is a better choice than a Word-derived format: it retains details of the layout and you don't need an editable format for archiving, and it's well defined with multiple existing implementations.
What the world needs is a good word processor that uses something like Docbook as its native format.
Can you give an example where patent trolls were cheered on for attacking Microsoft or Sony?
I'm saying you don't have a 10GB RAM disk, and your assertion that Vista makes RAM run poorly is pure conjecture. Let's say you somehow made a 10GB RAM disk, ignoring for the moment that you would have no RAM to run system, why you would go about installing any programs on a RAM disk is really amazingly stupid, the moment you restarted your computer the program would be annaliated.
That's "annihilated".
My i-RAM doesn't have that problem.
But you don't even need a battery backed SATA RAM drive to make a RAM disk that survives reboots. The Amiga Recoverable RAM Disk (RRD.DEVICE) hooked into the hardware boot vectors and did an integrity check at start, and so long as you hadn't lost power it would re-install itself with all its contents intact. You could even use it as your boot disk.
I've bought music based on ads linking to band sites.
Oh, man, if that holds up as a precedent... the implications are breathtaking.
I just felt a disturbance in the force, as if every patent troll in the country suddenly started hyperventilating and reaching for their blood pressure medication.
Hiro Protagonist?
Then might I recommend "The Book of the New Sun", starting with "The Shadow of the Torturer". See if you can count the number of words Severian uses to describe his profession...
Donaldson? The man who snorts the uncut Thesaurus most would have to be Gene Wolfe.
Oh yes, I text the ones I love too rather than make them grovel through voicemail. Text is polite, voicemail is evil and rude.
What does "[c]laims which are broad enough to read on obvious subject matter are unpatentable even though they also read on nonobvious subject matter." mean? It almost makes sense, but the term "read on" appears to be legal jargon, because it breaks /brain/lib/english_parser.so for me.
I assume by "iTunes" they mean "music player".
I didn't know iTunes had been kleenexed.
In hell, they make people like you take dictation via voice mail.
Voice mail is worse than talking to those boring people. I hate voice mail.
If I want to communicate with someone without calling them, I'll take text any day.
No, reading the Baroque cycle is like strapping yourself to a tree [...]
I didn't strap myself to that tree (or at least I cut myself loose before the end of the first volume). I'm talking about his science fiction, not his Swiss Family Robinson version of the 17th century.
Why not skip the idea of a separate browser "application" altogether and build web-rendering into something that resembles the "desktop,"
Microsoft did that in 1997. It was called "Active Desktop" and led to the biggest malware ecosystem ever.
The iPhone does not make that mistake. The browser is a separate application and the application, not the rendering engine, manages access between the sandbox and the outside world.
The iPhone costs three times as much as this device. If you subsidize it with a phone contract the list price is comparable, but then you have to add well over a thousand dollars per year for the contract.
Microsoft's problem in handhelds is Windows. They don't want the Windows Mobile based devices to become laptop replacements, because that would compete with Windows sales, but they want them to be recognizably Windows to both make development easier and to promote the brand.
Windows Mobile loses because Windows CE is just not reliable and solid enough to serve both the needs of a mobile phone and the needs of a general-purpose handheld. Palm didn't have this problem nearly as badly because PalmOS ran under a real time OS (AMX) that you couldn't get into from user applications... the whole Palm environment is just one task for AMX.
Take away the phone, and just worry about making a PDA, and you get a lot more freedom. The iPod Touch has really got more potential to benefit from iPhone apps than the iPhone, because it's not such a critical device. In the Pocket PC WinCE would have been fine if ActiveSync worked as well as Palm HotSync, so your ActiveSync repository served as a complete backup for everything on your Pocket PC... losing data every time I ran my battery flat was what drove me back to PalmOS for my PDA. So the Pocket PC loses because Microsoft didn't make it good enough to run standalone, and didn't make PC-side software good enough that you didn't care.
Going to a tablet, and you get even more freedom. Before "Tablet PC" there were Windows CE based clamshells and tablets that were quite capable, but Microsoft pretty much nuked them by loading the Pocket PC software down with restrictions (both technical and contractual) that meant the Windows CE based tablets were stuck with the previous generation of Windows Mobile software. Of course, they wanted their flagship product on the Tablet PC, not this stripped down embedded-only Windows CE.
I don't know if a browser-only tablet is a good solution, but a tablet is so far from the iPhone or Windows Mobile that trying to draw analogies between them is misleading at best... even if Microsoft hadn't continually undercut Windows Mobile to keep it from even potentially cannibalizing their flagship product.
The Diamond Age had the same problem.
Reading a Neal Stephenson novel is like strapping yourself into the back seat of a converted jet trainer to tour the Grand Canyon. For a lot of people, by the time they've gotten used to dodging pillars of rock at half the speed of sound and they're really enjoying the view the pilot flips over the rim and... that's all, tour's over.
I get used to the view pretty quick, and I've come to accept the endings, so I'll be picking up ANATHEM anyway.