A Status Update On GNU Hurd: Java, Debian, Money

[movieman]

I love how on the (incredibly rare) occasions when my Win7 video driver crashes, my desktop just flickers to black and back and then everything is back to running, no lost apps or anything.

I love how my Linux video drivers don't crash.

I used to see regular video driver crashes in Windows 7 using the driver that came with my laptop. I'd rather than just didn't crash than tried to bolt on a bodge to recover better when they do.

[Obscene_CNN]

Assuming HURD gets more than a single-digit number of dedicated developers, it at least has the chance of being better designed and better behaving (if not better performing) than Linux for hardcore users. However, I doubt it's ever going to be a project taken seriously by the people who matter (hardware vendors and consumers).

Ummm..... The one key flaw to the micro kernel design philosophy is performance. Too much time is wasted passing parameters between functions and different modules. Yes some researchers have shown that they can mitigate this to some extent but they can't eliminate it. In fact you can do some of the same tricks in a monolithic kernel to boost its performance further.

[XorEaxEax]

I love how my Linux video drivers don't crash.

Quoted for agreement.

I couldn't imagine working in an environment where any of my drivers routinely crash, but then again I'm guessing elanthis is just full of bullshit.

[V!NCENT]

Technically, HURD isn't a microkernel, but microkernel services on top of a microkernel.

Why am I saying this? Because there is also stalled effort to ditch Mach with its flaws for another L4 kernel, but that kernel development stalled due to the guy working on it being busy, so Mach is being used, but not developped for.

So why wasting resources on Mach? They are not; because HURD services use an interface with Mach, that when ported, being able to use all that work on other kernels if desired. So that why they call it services rather than a microkernel. It's also why they don't work on driver specific things like SATA directly.

[Remco]

Still, hybrid micro-kernels are pretty sweet.

I love how on the (incredibly rare) occasions when my Win7 video driver crashes, my desktop just flickers to black and back and then everything is back to running, no lost apps or anything. Upgrading a driver also requires no restarts. Linux... yeah, Linux. Crashes several times a month if you even think of maybe using your GPU for anything interesting; if even just X goes down all your apps are fu'd; and upgrading anything outside of a text editor usually requires replacing the kernel or half the low-level user-space libraries/daemons and rebooting. Not that you can actually get those updates until ~6 months from now when the distros deign to package up apps' new versions and throw them at you along with that cycle's flavor of desktop UI paradigm. Even though the Linux driver ABI problem and the distro package management problem makes rebooting a near necessity on interesting updates, since interesting updates only happen twice a year nobody notices.

Assuming HURD gets more than a single-digit number of dedicated developers, it at least has the chance of being better designed and better behaving (if not better performing) than Linux for hardcore users. However, I doubt it's ever going to be a project taken seriously by the people who matter (hardware vendors and consumers).

Attacking every single aspect of Linux in one short paragraph is not as effective as you might think.

Anyway, doesn't Linux have an early-stages form of driver rebooting?

[Remco]

Ummm..... The one key flaw to the micro kernel design philosophy is performance. Too much time is wasted passing parameters between functions and different modules. Yes some researchers have shown that they can mitigate this to some extent but they can't eliminate it. In fact you can do some of the same tricks in a monolithic kernel to boost its performance further.

Evidence please.
Thanks Michael.

[Ex-Cyber]

Ummm..... The one key flaw to the micro kernel design philosophy is performance. Too much time is wasted passing parameters between functions and different modules.

Well, "too much time wasted" is in the eye of the beholder. Someone running a massively multithreaded server workload might care whether a syscall takes 100 microseconds vs. 200 microseconds, but most users wouldn't be able to tell the difference if their lives depended on it.

[Caesar Tjalbo]

Assuming HURD gets more than a single-digit number of dedicated developers, it at least has the chance of being better designed and better behaving (if not better performing) than Linux for hardcore users.

I'm sure that as a hardcore desktop user, you'll do fine with Windows. How many developers it's going to take to make something of HURD I don't know, I'm not going to wait for it. Should you feel brave enough to throw Linux on your system, may I recommend a Long Term Stable or a rolling-release distro?

However, I doubt it's ever going to be a project taken seriously by the people who matter (hardware vendors and consumers).

Well, if GNU, Linux or QNX are any measure, there's a bright future for HURD.

[ultr]

What are you doing here, only complaining about Linux and praising Win 7? Don't you think there are then other places where you may pass your time more productively? Seriously.

As for HURD, I don't really see its usefulness. It's like Stallman never gave up on the idea he could release an OS entirely under his dream license.

There's no reason to give up that idea. Hurd is here for all who want to learn and to play, at least.
The (untainted/libre) Linux kernel is GPLv2 (not just his, but a dream license of us all) anyway, and there are also distros which actually entirely consist of free software only. Hurd is currently not even on the high priority projects list of FSF.

[Ibidem]

Still, hybrid micro-kernels are pretty sweet.

I love how on the (incredibly rare) occasions when my Win7 video driver crashes, my desktop just flickers to black and back and then everything is back to running, no lost apps or anything. Upgrading a driver also requires no restarts. Linux... yeah, Linux. Crashes several times a month if you even think of maybe using your GPU for anything interesting; if even just X goes down all your apps are fu'd; and upgrading anything outside of a text editor usually requires replacing the kernel or half the low-level user-space libraries/daemons and rebooting. Not that you can actually get those updates until ~6 months from now when the distros deign to package up apps' new versions and throw them at you along with that cycle's flavor of desktop UI paradigm. Even though the Linux driver ABI problem and the distro package management problem makes rebooting a near necessity on interesting updates, since interesting updates only happen twice a year nobody notices.

http://en.wikipedia.org/wiki/Xpra
I think that's what you missed.
Don't have many issues with graphics drivers here...
Just about any core Windows library, service, or kernel update takes a reboot; with Linux, libraries just mean running ldconfig (which dpkg/apt/... run automatically), and services are just "service xyz restart".
The kernel still takes a reboot, unless you know what you're doing (hint: ksplice)
The "interesting updates"--that would be like Windows Vista -> Seven.
You're confusing Canonical's representation of reality with how Linux actually works.