Uptime

[kataklinger]

You may want to see this:

http://en.uptime-project.net/page.php?p ... nt=windows

http://en.uptime-project.net/page.php?p ... tent=linux

http://en.uptime-project.net/page.php?p ... ent=others

http://en.uptime-project.net/page.php?p ... t=allhosts

[Eero Ränik]

I should start competing, I have a couple of useless machines sitting around here, and pretty stable power, though I should get a better UPS that'd keep the machine up for an hour or so it takes to get any problems fixed (not too often, but there's a chance that a storm or something like that would ruin the record attempt). Shouldn't be too hard to get such uptimes, other than that.

[Solar]

ACK.

Uptime is a pretty useless statement. I could fire up an ancient C64 and leave it running, and it would probably idle forever.

Security leak in the Linux kernel. Update the kernel (and reboot), or keep it running (and vulnerable)?

Kind of reminds me of the distributed.net "wars" of old. "Team Amiga" made it to the top 5. But does that tell anything about the calculation power of Amiga hardware, or rather about the ability of the Amiga community to rally under one flag?

[durand]

That's true. And, especially with those Windows machines, if it's been up more than a few weeks, then there's probably a few security holes exposed. Unless they completely hardened the machines 4 years ago and they're sitting behind several different kinds of firewalls with absolutely nothing exposed.

It would be interesting to run a scan on the top machines to see if they're vulnerable or not. Almost all security updates in Windows require a reboot.

Displaying uptime in these cases could be a case of disclosing information about your own vulnerable machines. In the case of the #1 Windows machine, it's acting as an FTP server. "I haven't updated in 4 years, 240 days, 8 hours and 50 minutes." Some administrator isn't doing their job...

[Solar]

I'm 100% with you, except for the bias towards Windows being more insecure.

Sifting through debian-security-announce archives 2005 lists two entries regarding the kernel: 2.4.27 (fixing 12 vulnerabilities) and 2.6.8 (27 vulnerabilities).

Both items carry the bottom line "We recommend that you upgrade your kernel package immediately and reboot the machine."

The top twelve Linux machines run on older kernels than that.

If you want to patch the kernel, you have to reboot. That's the same for either Windows or Linux.

And I still don't believe that being Open Source does per se mean you're more secure. It just means bugs might be fixed quicker if the supporting community is big enough.

Still requires a reboot, though.

[durand]

Oh yeah, totally. Fortunately, it's only kernel upgrades and stuff related to the kernel. The rest of userland can just be restarted. But I am negatilvey biased towards Windows, without any due justification, and I guess that showed I just installed their new patch released today related to that WMF thing that has everyone freaking out and that required a reboot...

Out of interest, the next version of the Windows kernel supports userland drivers. Have a look at http://channel9.msdn.com and the Going Deep series. They speak a lot about microkernel architecture and how Windows is heading towards that. It's interesting what those guys are doing over there. The one show has a kernel developer killing a device driver through the Task Manager and restarting it without any blue screen or problems.

So the increase of userland in Windows is possibly going to get rid of all those reboot for patches which are unrelated (directly) to the kernel and subsystems.

[durand]

Ah! And one of the things they're quite happy about is using their new technologies to analyze all the Windows binaries to discover what dependencies they all have.

One of the most interesting comments was that a modification in the standard shell (cmd.exe) can effect several hundred other binaries in the system because they rely on shell functionality in some way. Which, they mention, is counter-intuitive because the shell should be thought of independant and very little should rely on it to get it's job done. So, even an update to cmd.exe or some other arbitrary binary could require a reboot because of the level of dependancy in their system.

I'm very impressed with their channel9 series.

[Solar]

...and we won't see widespread use of userland drivers in Linux anytime soon, I guess, because then the Linux kernel maintainers could no longer force the hardware manufacturers to GPL their drivers, as they're trying to do.

Looking forward to the flamefests.

[Pype.Clicker]

(that reminds me of doing an apt-get upgrade )