I play console-style games like Madden, Spider-Man, and Tony Hawk.
I play first-person shooters, like Half-Life and Unreal Tournament.
I love hopping into Grand Theft Auto 3 for a taxi mission, but please don't take away the games that make the PC gaming experience so much richer than a console.
I would be very sad without Falcon 4.0, Grand Prix Legends, Morrowind, and Civilization 3.
These are truly escapist games that exercise more than your thumbs.
You have several options, ranging from $180 to $2,500.
$2,500 gets you RHEL AS with premium support, which includes 24/7 phone support for critical problems.
Most users will want RHEL ES with basic ($350) or standard ($700) support.
Red Hat doesn't make it easy to use Red Hat Enterprise Linux without paying for support.
You can get source RPMS from the FTP site, but you need Red Hat Enterprise Network with an RHEL entitlement to get ISOs and binary packages.
Someone with a subscription should be able to give you most of the packages, but possibly not the ISOs.
The RHEL subscription agreement is kind of scary, so I haven't been that interested.
I'm disappointed that Red Hat has only RHEL, for which you must pay every year, and RHL, which only gets a year of updates.
I'd consider paying a one-time fee of $100 - $200 for a Red Hat that had more than a year of updates.
This may be feasible if you're willing to build the RHEL errata from source, but that means you have to satisfy all the build requirements (e.g., -devel packages).
LVM makes it practical to use separate file systems, and it supports snapshots.
Since you've already got a separate/home file system, you may not want to bother migrating to LVM until your next full install.
Before LVM, I would just make/boot and / for maximum flexibility.
With LVM, I can make/boot,/,/home,/opt,/var,/tmp,/usr, and/usr/local.
/usr and/boot are mounted ro
/var and/tmp are mounted noatime
everything but / is mounted nodev
everything but / and/usr (and possibly/opt and/usr/local) are mounted nosuid, for what it's worth
If one of the file systems gets too full, I can resize it.
Ext3 has a nice tool called e2fsadm, but without on-line resize you may have to drop to single-user mode.
XFS and JFS support on-line resize, I think, but can only grow.
ReiserFS supports on-line resize (although the journal is a bit big for/tmp and/boot).
Snapshots make backups easier.
It's not a cure all, but it beats backing up an active file system.
Unfortunately, Red Hat has scarcely acknowledged a critical snapshot bug.
To get a clean snapshot, you need to temporarily lock the file system.
Red Hat's LVM doesn't do this, so you can't mount the snapshot.
(Snapshots are read only, so you can't replay the journal when the snapshot is dirty.)
You may be able to dump or dd the snapshot, but you can't mount it to use tar or cpio.
LVM makes me a little nervous, because it's one more thing that can go (and has gone) wrong.
All in all, though, I'd hate to do without it.
RTFA.
This is a new interview with Sterling Ball, published yesterday.
It's nice to see a status report, including the fact that the company is ditching its SCO systems because of the lawsuit.
I'd like to see some the guys involved with SMP or JFS or NUMA get together and *sue SCO.*
It's interesting that the slides mention XFS and NUMA, two SGI contributions.
Also, the BSD malloc() seems to have come from SGI, by way of HP.
Will SCO "revoke" the Irix, Tru64, and HP-UX licenses?
MSBlaster wasn't an embarrasment for MS, but for the lazy sysadmins who, with a month's prior notice and the patch to fix it, were still hobbled by the bug.
I'm using critical update notification on Windows 2000.
I installed a generic critical update the day before Blaster really took hold.
The next day, I had six new critical updates.
That same day, Windows Update on three Windows XP systems showed no updates.
when I ran Windows Update again in the afternoon, there were twenty critical updates.
If the patch has really been available for months, then Windows Update is severely broken.
If it doesn't work when I'm actively using it, why would I want it to be automatic?
The comparison to the GNU FTP site is specious.
On the one hand, a million computers were compromised by a worm; on the other, one FTP server was compromised by an insider.
The clearest explanation is in the RHEL subscription agreement.
Premium and standard subscriptions entitle you to "new versions of the applicable Software, if any, released during the subscription period."
Basic subscriptions do not.
I notice that RHEL ES standard is temporarily reduced from $800 to $700, although that's still twice as much as RHEL ES basic.
Actually, five years of support would cost $4,000 ($800 per year).
Red Hat promises to release errata for five years, but it may be difficult to get them if you let your support contract lapse.
Hire 3 people, and within 2 years you're within the $350K range.
And, all this is assuming the people you hire are as skilled with Linux as a Red Hat support professional.
You're going to need skilled sysadmins even if you pay $350,000 for RHEL.
Red Hat doesn't sell a turn-key, sysadmin-in-the-box solution like, say, Sun or EMC.
(E.g., support tech. shows up at your door and tells you one of your hard drives is about to fail.)
If they did, it would probably cost even more for the submitter's scenario.
They're moving in that direction with products like Enterprise Network.
Its not like... your apps only run on a single vendor's platform.
So you can get Oracle, Sybase, IBM, BEA, Alias, Pixar, Side Effects, and Softimage to support your Debian, Mandrake, and Gentoo systems?
Red Hat is free to charge $200 - $2,500 per system (although I'm still uneasy about the EULA that allows audits, etc.).
Proprietary software vendors are free to support only expensive distributions.
At the end of the day, a customer can be forgiven for wondering where the "free" went.
You're free to choose RHEL or SLES, but if you exercise the freedom to modify the programs contained therein, you'll be running an unsupported configuration.
I'm wondering... don't people use firewalls, at ALL?
Laptops often make an end run around firewalls.
I attended a Windows 2003 seminar at a Microsoft branch a few months ago.
We lost our Internet connection twice, because someone brought in a Slammer-infected laptop.
It sucked, because Google is more effective than MS help files.
We had similar situations at work with Nimbda and Slammer.
I think Nimbda came in over VPN.
Well after Slammer's peak, a consultant brought it in on a laptop.
If you want a flexible work environment, a firewall is not enough.
Yes, performance is acceptable even on a lowly Pentium Pro 200 over the network.
So far, I've built pretty simple prototypes.
I'm not sure how well libglade and Python will scale.
Libglade loads an XML file at run time that was created with GLADE, the user interface builder.
I wonder whether the load times will increase appreciably as the application gets more complex.
Python isn't the fastest language and doesn't have the best threading support.
I'm not sure how it will deal with rapid-fire signals, like mouse motion events and time outs (say, for animating a spinner logo).
I don't really anticipate needing to handle these in my current project, but I'd like to test them.
If Python doesn't scale well, I can switch from PyGTK to gtkmm or GTK+.
If libglade doesn't scale well, I can switch to generated code.
The only two that can even come close [to Qt] are GNOME and wxWindows, but really who wants to write in plain C?
I don't touch C, except when I'm submitting patches to the GNOME libraries.
I use PyGTK, the Python binding.
James Henstridge and friends have done a bang-up job.
GTK+, GLADE, libglade, PyGTK, and PyGNOME are a great set of tools.
My biggest concern with GNOME is the apparent lack of a commitment to maintain backwards compatibility.
This has hurt Galeon and GnuCash, I think, to say nothing of the time wasted porting other projects from GNOME 1 to GNOME 2.
Maybe they'll have this worked out when it's a little more mature.
While I haven't spent as much time with PyQt, it didn't blow me away the way PyGTK did.
It uses old-style, rather than new-style, classes.
PyKDE is a separate project: while the latest release of PyQt is 3.7, you have to stay at 3.5 if you want to use PyKDE.
Finally, the 250-GBP commercial license fee is ridiculous.
I guess they have to pay for the Qt commercial license, somehow.
Three words: terms of service.
You've seen (or, more likely, ignored) them in the fine print of corporate web sites.
I propose that mirror sites include in their login banners and robots.txt file a bit about fees for "consulting" and bandwidth costs due to unauthorized use of the site and any correspondence arising from such use.
If the lawyer refuses to pay the bill for the research he requested, inform the ethics committee of the local branch of the ABA about his "good faith" claim.
There are some legal procedures, like rule 11, that are never invoked, because no one would ever be that stupid.
There has to be a rule somewhere to stop these DMCA letters, because they are that stupid.
If SCO's users can't develop software for their chosen platform anymore, then they will likely choose another platform
The availability of GCC and other free software on OpenServer and UnixWare may make it easier to eventually migrate off of those platforms.
If a user has a compiler, he can build Apache, MySQL, and PHP, in preparation to migrate from SCAMP to LAMP.
As tempting as it is to excommunicate a platform for political reasons, it's a bad idea.
OpenServer and UnixWare support may eventually die due to bit rot, but don't remove it out of spite.
On my machine the CPU fan drowns out the noise of any PSU
The power supply fan has always been louder than the CPU fan in my computers.
I had a generic 300 W PSU that I thought was too loud.
I replaced it with a PC Power & Cooling "Silencer" that was just as loud.
I later got an "ultra-quiet" QTechnology PSU from Quiet PC.
Little, if any, improvement.
Noise is important to some people, and it appears to be a difficult problem to solve.
Not too long ago, 300 W PSUs were considered extravagant, but it's easy to overtax one with a few disks, especially if the PSU isn't up to spec.
One of my computers rebooted sporadically, until I replaced the PSU.
If Linux is to survive, it is imperative that its license, the obnoxious GPL, is tested in court.
It's unfair to call the GPL obnoxious, in the context of this lawsuit.
Proprietary software licenses and contracts are responsible for this mess, including the amendment that SCO produced, showing that it owns the System V copyrights, and the clause that Novell invoked, forbidding SCO from terminating IBM's license.
Microsoft executives have labeled the GPL an un-American, anti-business cancer.
The companies improving GPLed projects are too numerous to list, but include fierce competitors, such as IBM, HP, Sun, SGI, Novell, Oracle, Red Hat, and SuSE.
Cooperation like that is inconceivable with traditional cross licensing.
(Remember Taligent or Monterey?)
Most computer users are utterly and completely retarded concerning anything related to computers....
it's going to be easier to solve stuff later on as the general population slowly becomes more tech-savvy. Still, a few good regulations regarding the teaching about computer usage might be nice
I recognize that some of the tasks we perform with computers are inherently complex.
However, I think the emphasis must always be on improving the user interface so that a lifetime of training is not a prerequisite to, say, burning a CD.
Donald Norman'sDesign of Everyday Things provides an insightful perspective on this issue.
For example, a door that requires an instruction manual, however short ("Push" and "Pull"), is poorly designed.
His Invisible Computer is an interesting comparison of computers to motors.
Motors used to be expensive, so a household would rarely own more than one, to which several proprietary peripherals, like sewing machines, could be attached.
Today, of course, we wouldn't dream of buying a blender that required a separate motor.
Likewise, computers are making the transition from expensive mainframes ("I think there is a world market for maybe five computers") to emedded devices.
my standard debug procedure, had he acted like a dumb user would have been far faster
I can accept that.
I can't accept a script that includes unnecessarily destructive steps, like reformatting the hard drive.
Also, if the script includes "go-away" steps (something that takes more than a couple of minutes), I shouldn't have to wait in a full phone queue when I call back.
Red Hat is very careful to keep packages clean for architectures other than x86.
IA-64 support was one of the reasons for the controversial GCC 2.96 fork.
future development of Evolution as an open source product will be called into question
I was wondering the same thing, myself.
I predict that Evolution will go the way of Nautilus.
The Evolution developers will be reassigned to do a native GroupWise port.
A patchwork of GNOME developers will keep Evolution on life support (i.e., adapting it to each backwards-incompatible revision of GTK+).
I don't miss many features in Evolution, so this would be an acceptable scenario.
If anything, I'm more worried that Evolution will undergo a Galeon-style rewrite that breaks existing features or changes the user interface.
I play console-style games like Madden, Spider-Man, and Tony Hawk. I play first-person shooters, like Half-Life and Unreal Tournament. I love hopping into Grand Theft Auto 3 for a taxi mission, but please don't take away the games that make the PC gaming experience so much richer than a console. I would be very sad without Falcon 4.0, Grand Prix Legends, Morrowind, and Civilization 3. These are truly escapist games that exercise more than your thumbs.
You have several options, ranging from $180 to $2,500. $2,500 gets you RHEL AS with premium support, which includes 24/7 phone support for critical problems. Most users will want RHEL ES with basic ($350) or standard ($700) support.
Red Hat doesn't make it easy to use Red Hat Enterprise Linux without paying for support. You can get source RPMS from the FTP site, but you need Red Hat Enterprise Network with an RHEL entitlement to get ISOs and binary packages.
Someone with a subscription should be able to give you most of the packages, but possibly not the ISOs. The RHEL subscription agreement is kind of scary, so I haven't been that interested.
I'm disappointed that Red Hat has only RHEL, for which you must pay every year, and RHL, which only gets a year of updates. I'd consider paying a one-time fee of $100 - $200 for a Red Hat that had more than a year of updates. This may be feasible if you're willing to build the RHEL errata from source, but that means you have to satisfy all the build requirements (e.g., -devel packages).
LVM makes it practical to use separate file systems, and it supports snapshots. Since you've already got a separate /home file system, you may not want to bother migrating to LVM until your next full install.
Before LVM, I would just make /boot and / for maximum flexibility.
With LVM, I can make /boot, /, /home, /opt, /var, /tmp, /usr, and /usr/local.
- /usr and
/boot are mounted ro
- /var and
/tmp are mounted noatime
- everything but / is mounted nodev
- everything but / and
/usr (and possibly /opt and /usr/local) are mounted nosuid, for what it's worth
If one of the file systems gets too full, I can resize it. Ext3 has a nice tool called e2fsadm, but without on-line resize you may have to drop to single-user mode. XFS and JFS support on-line resize, I think, but can only grow. ReiserFS supports on-line resize (although the journal is a bit big forSnapshots make backups easier. It's not a cure all, but it beats backing up an active file system. Unfortunately, Red Hat has scarcely acknowledged a critical snapshot bug. To get a clean snapshot, you need to temporarily lock the file system. Red Hat's LVM doesn't do this, so you can't mount the snapshot. (Snapshots are read only, so you can't replay the journal when the snapshot is dirty.) You may be able to dump or dd the snapshot, but you can't mount it to use tar or cpio.
LVM makes me a little nervous, because it's one more thing that can go (and has gone) wrong. All in all, though, I'd hate to do without it.
RTFA. This is a new interview with Sterling Ball, published yesterday. It's nice to see a status report, including the fact that the company is ditching its SCO systems because of the lawsuit.
When I got a voice mail from Reg last year, he pronounced it S-C-O. That was the first time I'd ever heard it spelled out.
It's interesting that the slides mention XFS and NUMA, two SGI contributions. Also, the BSD malloc() seems to have come from SGI, by way of HP. Will SCO "revoke" the Irix, Tru64, and HP-UX licenses?
I'm using critical update notification on Windows 2000. I installed a generic critical update the day before Blaster really took hold. The next day, I had six new critical updates.
That same day, Windows Update on three Windows XP systems showed no updates. when I ran Windows Update again in the afternoon, there were twenty critical updates.
If the patch has really been available for months, then Windows Update is severely broken. If it doesn't work when I'm actively using it, why would I want it to be automatic?
The comparison to the GNU FTP site is specious. On the one hand, a million computers were compromised by a worm; on the other, one FTP server was compromised by an insider.
The clearest explanation is in the RHEL subscription agreement. Premium and standard subscriptions entitle you to "new versions of the applicable Software, if any, released during the subscription period." Basic subscriptions do not.
I notice that RHEL ES standard is temporarily reduced from $800 to $700, although that's still twice as much as RHEL ES basic.
Actually, five years of support would cost $4,000 ($800 per year). Red Hat promises to release errata for five years, but it may be difficult to get them if you let your support contract lapse.
You're going to need skilled sysadmins even if you pay $350,000 for RHEL. Red Hat doesn't sell a turn-key, sysadmin-in-the-box solution like, say, Sun or EMC. (E.g., support tech. shows up at your door and tells you one of your hard drives is about to fail.) If they did, it would probably cost even more for the submitter's scenario. They're moving in that direction with products like Enterprise Network.
So you can get Oracle, Sybase, IBM, BEA, Alias, Pixar, Side Effects, and Softimage to support your Debian, Mandrake, and Gentoo systems?
Red Hat is free to charge $200 - $2,500 per system (although I'm still uneasy about the EULA that allows audits, etc.). Proprietary software vendors are free to support only expensive distributions. At the end of the day, a customer can be forgiven for wondering where the "free" went. You're free to choose RHEL or SLES, but if you exercise the freedom to modify the programs contained therein, you'll be running an unsupported configuration.
That's news to Oracle: What Distributions of Linux Does Oracle Support?
Surprise, surprise. After years of dumping to annihilate the competition, the prices go back up.
Laptops often make an end run around firewalls. I attended a Windows 2003 seminar at a Microsoft branch a few months ago. We lost our Internet connection twice, because someone brought in a Slammer-infected laptop. It sucked, because Google is more effective than MS help files.
We had similar situations at work with Nimbda and Slammer. I think Nimbda came in over VPN. Well after Slammer's peak, a consultant brought it in on a laptop.
If you want a flexible work environment, a firewall is not enough.
Libglade loads an XML file at run time that was created with GLADE, the user interface builder. I wonder whether the load times will increase appreciably as the application gets more complex.
Python isn't the fastest language and doesn't have the best threading support. I'm not sure how it will deal with rapid-fire signals, like mouse motion events and time outs (say, for animating a spinner logo). I don't really anticipate needing to handle these in my current project, but I'd like to test them.
If Python doesn't scale well, I can switch from PyGTK to gtkmm or GTK+. If libglade doesn't scale well, I can switch to generated code.
I don't touch C, except when I'm submitting patches to the GNOME libraries. I use PyGTK, the Python binding. James Henstridge and friends have done a bang-up job. GTK+, GLADE, libglade, PyGTK, and PyGNOME are a great set of tools. My biggest concern with GNOME is the apparent lack of a commitment to maintain backwards compatibility. This has hurt Galeon and GnuCash, I think, to say nothing of the time wasted porting other projects from GNOME 1 to GNOME 2. Maybe they'll have this worked out when it's a little more mature.
While I haven't spent as much time with PyQt, it didn't blow me away the way PyGTK did. It uses old-style, rather than new-style, classes. PyKDE is a separate project: while the latest release of PyQt is 3.7, you have to stay at 3.5 if you want to use PyKDE. Finally, the 250-GBP commercial license fee is ridiculous. I guess they have to pay for the Qt commercial license, somehow.
Three words: terms of service. You've seen (or, more likely, ignored) them in the fine print of corporate web sites. I propose that mirror sites include in their login banners and robots.txt file a bit about fees for "consulting" and bandwidth costs due to unauthorized use of the site and any correspondence arising from such use.
If the lawyer refuses to pay the bill for the research he requested, inform the ethics committee of the local branch of the ABA about his "good faith" claim.
There are some legal procedures, like rule 11, that are never invoked, because no one would ever be that stupid. There has to be a rule somewhere to stop these DMCA letters, because they are that stupid.
The availability of GCC and other free software on OpenServer and UnixWare may make it easier to eventually migrate off of those platforms. If a user has a compiler, he can build Apache, MySQL, and PHP, in preparation to migrate from SCAMP to LAMP.
As tempting as it is to excommunicate a platform for political reasons, it's a bad idea. OpenServer and UnixWare support may eventually die due to bit rot, but don't remove it out of spite.
The power supply fan has always been louder than the CPU fan in my computers. I had a generic 300 W PSU that I thought was too loud. I replaced it with a PC Power & Cooling "Silencer" that was just as loud. I later got an "ultra-quiet" QTechnology PSU from Quiet PC. Little, if any, improvement. Noise is important to some people, and it appears to be a difficult problem to solve.
Not too long ago, 300 W PSUs were considered extravagant, but it's easy to overtax one with a few disks, especially if the PSU isn't up to spec. One of my computers rebooted sporadically, until I replaced the PSU.
It's unfair to call the GPL obnoxious, in the context of this lawsuit. Proprietary software licenses and contracts are responsible for this mess, including the amendment that SCO produced, showing that it owns the System V copyrights, and the clause that Novell invoked, forbidding SCO from terminating IBM's license.
Microsoft executives have labeled the GPL an un-American, anti-business cancer. The companies improving GPLed projects are too numerous to list, but include fierce competitors, such as IBM, HP, Sun, SGI, Novell, Oracle, Red Hat, and SuSE. Cooperation like that is inconceivable with traditional cross licensing. (Remember Taligent or Monterey?)
I recognize that some of the tasks we perform with computers are inherently complex. However, I think the emphasis must always be on improving the user interface so that a lifetime of training is not a prerequisite to, say, burning a CD. Donald Norman's Design of Everyday Things provides an insightful perspective on this issue. For example, a door that requires an instruction manual, however short ("Push" and "Pull"), is poorly designed.
His Invisible Computer is an interesting comparison of computers to motors. Motors used to be expensive, so a household would rarely own more than one, to which several proprietary peripherals, like sewing machines, could be attached. Today, of course, we wouldn't dream of buying a blender that required a separate motor. Likewise, computers are making the transition from expensive mainframes ("I think there is a world market for maybe five computers") to emedded devices.
I can accept that. I can't accept a script that includes unnecessarily destructive steps, like reformatting the hard drive. Also, if the script includes "go-away" steps (something that takes more than a couple of minutes), I shouldn't have to wait in a full phone queue when I call back.
Red Hat Enterprise Linux 2.1 supports x86 and IA-64:
Red Hat Enterprise Linux 3.0 is in beta for x86, x86-64, IA-64, pSeries, and iSeries:
Red Hat is very careful to keep packages clean for architectures other than x86. IA-64 support was one of the reasons for the controversial GCC 2.96 fork.I was wondering the same thing, myself. I predict that Evolution will go the way of Nautilus. The Evolution developers will be reassigned to do a native GroupWise port. A patchwork of GNOME developers will keep Evolution on life support (i.e., adapting it to each backwards-incompatible revision of GTK+).
I don't miss many features in Evolution, so this would be an acceptable scenario. If anything, I'm more worried that Evolution will undergo a Galeon-style rewrite that breaks existing features or changes the user interface.