Microsoft never said that, infact they have articles on their site describing the bad effects of fragmentation. They probably didn't ship one with NT4 to push people to buy one from a 3rd party.
And how do you restart a daemon that doesn't handle SIGHUP with BSD init? With SysV I just type/etc/init.d/daemon restart.
SysV can be a little overwhelming at first, but to me BSD was just too limiting because there's no system for individual startup scripts and I'm not going to waste my time creating one when SysV works so well.
Re:My experiences with Gentoo
on
Gentoo Reviewed
·
· Score: 1
The only thing I might consider a flaw in apt's depdency handling is the fact that it doesn't honor packages in Recommends or Suggests, it only installs packages in Depends. But I think this is being changed and I usually use dselect anyway, which handles Recommends and Suggests just fine.
And if I ever got the urge to compile something, heaven forbid, I can just use apt-build.
Re:My experiences with Gentoo
on
Gentoo Reviewed
·
· Score: 1
Even if SuSe uses an initrd on their kernels that doesn't mean you have to with your custom ones. Almost every distro uses an initrd by default now so that everything can be a module and loaded only if necessary, but that doesn't affect what you do after the install is complete.
You will never see any sort of broken dependencies with the official Debian repositories, if you add 3rd party APT sources you're taking your chances.
Personally I've been using Debian on my workstation for over 4 years and I havn't had a problem that wasn't my own fault or didn't have the package fixed in a day or so and I've been running unstable the whole time.
Everyone uses FAT on them so they will work under Windows, but it's a very crappy filesystem, not supported on all OSes, and Windows is the only platform that has the tools to work with FAT filesystems properly (Defrag, Chkdsk, etc).
What OS doesn't support FAT? I'm pretty sure even IOS on some routers and SRM on Alphas reads FAT floppies for firmware updates.
dosfsck (8) - check and repair MS-DOS file systems
I don't see a defrag util, but that's not a show stopper IMO.
The ONLY reason UFS isn't everywhere, is because Windows doesn't support it at all.
So why don't someone write a Windows driver for UFS? Even a driver that can be installed after the main OS would be a decent, so Windows can run off of NTFS but data can be stored and shared on a UFS partition. Maybe it's because there's so many variants, the UFS driver for Linux supports 7 different types of UFS. I don't think UFS will become universal as long as it's a name that refers to multiple types of filesystems.
Sometimes I wish there was the equivalent of Windows Update for Linux
apt-get update apt-get upgrade
I don't run Debian's precompiled kernels though so I don't know what the patch/release policy on them is, but for all userland things it's better than WU.
I would say it's more consistency than stability, there's no chance of an 'apt-get upgrade' on woody installing the Samba 3.x cvs that's currently in sid which would be a huge 'upgrade'. Security fixes are backported but major version upgrades generally don't happen. I personally run sid on my workstation and rarely have problems, usually just a broken package that's fixes in a few hours.
You don't get crashes at all with FreeBSD.
Except for that minor bug that caused FreeBSD to panic on Alphas when the dhcp client was used. It's really frustrating trying to install FreeBSD when the thing keeps panicing just on the install.
The difference would be support. If I have a problem with Oracle on FreeBSD they're going to hang up on me and no matter how much the support may suck it's still an issue for most businesses. (note: I have no idea how good or bad Oracle support is, I just know most of my run-ins with support have been bad so I don't get my hopes up)
No differences other than the fact that they don't usually use x86 CPUs, have a RT process scheduler, run from nvram, etc.
The big problem would be getting a ARM (or whatever the embedded arch is) binary to run on an Intel CPU, you'd need an emulator of sorts and it would probably be too slow anyway.
And an IIS vulnerability has nothing to do with the NT Kernel, only that it's run on it. So if an IIS or SQL worm can be called a 'windows' bug, then an Apache or mySql one can be called a 'linux' bug, it's only fair.
Not really. IIS and MS SQL only run on Windows, Apache and MySQL run on quite a few OSes other than Linux. I personally wouldn't call a MS SQL exploit a "Windows bug" but you'd still be hard pressed to find another environment that it'll cause problems in.
Microsoft no doubt has a team of employees who look for vulnerabilities
If that's true, that team is obviously under no obligation to tell anyone when they find them, not even their employers =)
Open source is great because it can be reviewed by it's peers. Wonderful. But who reviews it? Do you? I know I haven't pored through the source for the SAMBA tarball I just installed on one of my boxes looking for holes.
Not everyone reads the code, not everyone should. But the fact that you can is a big advantage (I actually had to change something in the Samba src once). You (and most people) rely on the Samba developers to review each other, just like when I see a patch posted on lkml a handfull of people flock to it and say whether the patch looks correct or does stupid things, I would expect the same to happen with Samba.
actually they've shipped 64-bit support for years, back in the Alpha days and in windows 2000 and XP (they also have a guide here [microsoft.com]).
The article you mention talks about Windows Server 2003, not Win2K or XP. I'm not saying Win64 isn't real, but it's more recent than you're saying. The NT ports to Alpha were all 32-bit, not 64-bit.
Yeah, and Intel and MS have been 64 bit for years now as well. THEY KNOW that 64 bit is no big deal because they've already been there.
When has MS been there? The NT ports to 64-bit arches like Alpha have been 32-bit, if they've been there why did it take so long to get Win2K RTM on Itanium? HP-UX and Linux both were done marginally quicker than Win2K, because they had already been there.
Are you sure about that? Since each process gets its own address space, I would assume that it also gets its own 2GB limit. (*some* of it is shared, like system dlls)
The 2G limit is per-process. And the shared libraries don't even come into play, each library mapped into a process's address space takes up the same amount per process even though it's only in memory one time. So if loading MFC.dll takes up 10M of virtual address space, it takes up 10M in each app's space that needs it, even though it's only using 10M physical memory.
Slashdot Mirror
I'm using XFS (have been for a few years) and I've never used a distribution kernel on this box.
All I'm FORCED to do, is apply the XFS patches from the SGI page, which is a minor step.
HOST $dir login.com
Directory DEV:[DIRECTORY]
LOGIN.COM;10 LOGIN.COM;9 LOGIN.COM;8
Total of 3 files.
Hasn't stopped them before. They've changed the on-disk format in NTFS in NT 4 SP4, Win2K and WinXP.
xfs_repair
But it's not recommend to be run on a mounted filesystem, even one mounted ro IIRC.
Microsoft never said that, infact they have articles on their site describing the bad effects of fragmentation. They probably didn't ship one with NT4 to push people to buy one from a 3rd party.
Before you jump all over SGI for no reason, read the man page for xfs_repair, it's been in the Linux release for as long as I can remember.
fsck.xfs is only setup to always return 0 for compatiblity with startup scripts that fail without a fsck program.
And how do you restart a daemon that doesn't handle SIGHUP with BSD init? With SysV I just type /etc/init.d/daemon restart.
SysV can be a little overwhelming at first, but to me BSD was just too limiting because there's no system for individual startup scripts and I'm not going to waste my time creating one when SysV works so well.
The only thing I might consider a flaw in apt's depdency handling is the fact that it doesn't honor packages in Recommends or Suggests, it only installs packages in Depends. But I think this is being changed and I usually use dselect anyway, which handles Recommends and Suggests just fine.
And if I ever got the urge to compile something, heaven forbid, I can just use apt-build.
Even if SuSe uses an initrd on their kernels that doesn't mean you have to with your custom ones. Almost every distro uses an initrd by default now so that everything can be a module and loaded only if necessary, but that doesn't affect what you do after the install is complete.
You will never see any sort of broken dependencies with the official Debian repositories, if you add 3rd party APT sources you're taking your chances.
Personally I've been using Debian on my workstation for over 4 years and I havn't had a problem that wasn't my own fault or didn't have the package fixed in a day or so and I've been running unstable the whole time.
What OS doesn't support FAT? I'm pretty sure even IOS on some routers and SRM on Alphas reads FAT floppies for firmware updates.
dosfsck (8) - check and repair MS-DOS file systems
I don't see a defrag util, but that's not a show stopper IMO.
The ONLY reason UFS isn't everywhere, is because Windows doesn't support it at all.
So why don't someone write a Windows driver for UFS? Even a driver that can be installed after the main OS would be a decent, so Windows can run off of NTFS but data can be stored and shared on a UFS partition. Maybe it's because there's so many variants, the UFS driver for Linux supports 7 different types of UFS. I don't think UFS will become universal as long as it's a name that refers to multiple types of filesystems.
apt-get update
apt-get upgrade
I don't run Debian's precompiled kernels though so I don't know what the patch/release policy on them is, but for all userland things it's better than WU.
Whenever you see a statement like this it's usually referring to the filesystem ACLs where it says "security model".
And the piracy protection in XP is activation, not registration, registration is very different.
Because the BSD userland and ports suck compared to the GNU userland and apt/dpkg.
You don't get crashes at all with FreeBSD.
Except for that minor bug that caused FreeBSD to panic on Alphas when the dhcp client was used. It's really frustrating trying to install FreeBSD when the thing keeps panicing just on the install.
IIRC apt-build has a 'world' option to rebuild every package on the system, although I can't really see why you'd want to.
If you really think Linux is more of a PITA than a BSD you havn't tried Debian.
I believe SuperMicro plans on releasing server and workstation boards for the Opteron and the workstation boards should have an AGP slot.
Intel freely admits their 32-bit emulation is crap, they're not aiming for the desktop with IA-64 so it's not a big deal.
The difference would be support. If I have a problem with Oracle on FreeBSD they're going to hang up on me and no matter how much the support may suck it's still an issue for most businesses. (note: I have no idea how good or bad Oracle support is, I just know most of my run-ins with support have been bad so I don't get my hopes up)
No differences other than the fact that they don't usually use x86 CPUs, have a RT process scheduler, run from nvram, etc.
The big problem would be getting a ARM (or whatever the embedded arch is) binary to run on an Intel CPU, you'd need an emulator of sorts and it would probably be too slow anyway.
Not really. IIS and MS SQL only run on Windows, Apache and MySQL run on quite a few OSes other than Linux. I personally wouldn't call a MS SQL exploit a "Windows bug" but you'd still be hard pressed to find another environment that it'll cause problems in.
Microsoft no doubt has a team of employees who look for vulnerabilities
If that's true, that team is obviously under no obligation to tell anyone when they find them, not even their employers =)
Open source is great because it can be reviewed by it's peers. Wonderful. But who reviews it? Do you? I know I haven't pored through the source for the SAMBA tarball I just installed on one of my boxes looking for holes.
Not everyone reads the code, not everyone should. But the fact that you can is a big advantage (I actually had to change something in the Samba src once). You (and most people) rely on the Samba developers to review each other, just like when I see a patch posted on lkml a handfull of people flock to it and say whether the patch looks correct or does stupid things, I would expect the same to happen with Samba.
The article you mention talks about Windows Server 2003, not Win2K or XP.
I'm not saying Win64 isn't real, but it's more recent than you're saying. The NT ports to Alpha were all 32-bit, not 64-bit.
When has MS been there? The NT ports to 64-bit arches like Alpha have been 32-bit, if they've been there why did it take so long to get Win2K RTM on Itanium? HP-UX and Linux both were done marginally quicker than Win2K, because they had already been there.
Intel has already licensed x86-64 "just in case", from what I hear.
The 2G limit is per-process. And the shared libraries don't even come into play, each library mapped into a process's address space takes up the same amount per process even though it's only in memory one time. So if loading MFC.dll takes up 10M of virtual address space, it takes up 10M in each app's space that needs it, even though it's only using 10M physical memory.