Slashdot Mirror

Robert has done plenty of useful work, but it was George AnzigerAnzinger who developed the Linux preemption patch. Robert picked it up, maintained it and got it merged. The credits to George seemed to have gotten lost somewhere in that process.

Credit where credit is due please.

The GP claimed Microsoft had sabotaged ACPI -- I simply was pointing out that ACPI is in fact broken, and not functional as per the specification as it seemed you believed.

The presentation PDFs are available in big chunks (you'll have to search) from the proceedings page.

You need to have a conversation with those who try to make ACPI work to spec in Linux then. Its an open spec, Linux is open source, now ask them why power management doesn't work as advertised. It was discussed heavily at the OLS two years ago as I recall. There's PDFs on the site if you want to read the presentations.

When I was at OLS a couple years ago there was a good presentation on how much power management sucked in its actual implementation by hardware vendors and why it was such a problem to implement ACPI properly in Linux. If you go on the site above you should be able to find the PDFs of the presentation.

My suggestion would be to find someone who's pretty savvy in the area you're aimed at, and hire him or her (OK, let's face it, "him"...)

The gender ratio is pretty extreme, but it's not 100%--there *are* expert female kernel hackers.

I'm just guessing that finding a kernel guru willing to give up a month of Saturday afternoons at $300 a session will be easier than finding "Linux Kernel for Experts" at the downtown Learning Annex.

Personal tutoring is a pretty expensive way to get an education, especially if it's in a fast-moving field whose experts are in demand for other work.

Off the top of my head:

• Volunteer, and find a problem to work on.
• Find someone to hire you to do kernel work, or some other way to work with people doing what you want to do.
• Have you considered grad school? There's places where you could get a degree taking operating systems classes and hacking the kernel for your dissertation. And when you're done you'll probably have an easier time with the previous item.
• Local user groups and universities might be good places to meet up with people who share your interests. Maybe start a study group to learn some kernel subsystem together?
• Conferences: OLS and linux.conf.au are fun. The OLS papers at least are on line if you can't go.
• Mailing lists, irc channels--look at kernelnewbies, etc.
• Books: Linux Device Drivers, Understanding the Linux Kernel, Robert Love's book.
• Read the code!

And if people have told you that "the best method would be to dig into the kernel myself",... actually, in the end, it's probably the *only* way. There's a certain point in your study of any field where you just run out of "courses". That's good. It means you're ready to do real work, because you're at the point where people are still busy doing the work and figuring stuff out, and haven't yet figured out how to break it down into manageable chunks and explain it in a logical order, which is a great deal of work in and of itself.

• Microsoft - Conspiracy theorists find a way to blame Microsoft for every problem in Linux. This time they point out when Microsoft decided to use a vendor's chip in the XBox consoles or chipset vendors puts DirectX 8.0 support you don't get specs any more.
  
• ??? - Patents and fear of competitors or patent scumsucking companies bringing infringement. Vendors claim releasing chipset docs to the public may make it easier for these things to be found; however, most X.org developers have no problem signing suitable NDAs .
  
• Profit - Graphics card manufacturing is a very competitive industry, especially in the high-end gaming, 3-6 month development cycle, grind-out-
  as-many-different-cards-as-you-can. Quake 3 speeds are spotted in binary drivers any way and it doesn't explain fglrx which are some of the most unsuitable drivers for gaming on Linux.

OK, now I've even searched the referenced (but not linked) article, which someone helpfully linked to above, and I've STILL been unable to locate this Blizzard. Does that quote have ANYTHING to do with the story? If I were to throw-up in the Story Submission box, would a Slashdot editor clean it up, or submit the story as-is?

So you dont have to dig...

"Why user space sucks" is at:

Pages 441-449 of http://www.linuxsymposium.org/2006/linuxsymposium_ procv1.pdf

You know, in addition to reading that eWeek article you posted, I also read the paper presented at the 2006 Linux Symposium, which the eWeek article referenced. I know how the thing works now!

Nevertheless, I'll read the articles you linked, and use them to refute your points. From your first link:

To reduce the maintenance burden as much as possible, while still allowing the implementation to accommodate changes, the design provides a stable ABI with semantic invariants.

"ABI" stands for "Application Binary Interface." In other words, it allows linking a binary blob into the kernel.

As I said before, a stable ABI is what the kernel maintainers have been denying to nVidia and ATi -- and for good reason! Why should VMWare be treated differently?

the current stable kernel developer says he agrees with VMware's methodology (a common VM layer rather than a specific one to Xen)

Here's the exact quote (from your second article):

"From a high-level design perspective, I think that VMware's point is a good one, and that a general kernel-to-virtual machine interface is a better thing than a Xen-only one," Morton said.

That's not the same thing as "agree[ing] with VMWare's methodology." "VMWare's point" could equally well apply to having a general API, which is not the same as what's actually being done.

Almost 1,000 pages of very interesting whitepapers from the event can be found in the first and second PDFs.

Almost 1,000 pages of very interesting whitepapers from the event can be found in the first and second PDFs.

HTTP-FUSE Xenoppix is presented at Ottawa Linux Symposium 2006.

http://www.linuxsymposium.org/2006/view_abstract.p hp?content_key=76

• Theory says hardware is faster than software; in this case, theory is wrong. Workstation 5.5 has a full, optimized VT implementation. It's no faster than the normal binary translation approach, and in significant cases does worse. (Full details are at an academic conference in October; the paper itself is hard to get, sorry.). The short version is, switching in and out of VT mode is so expensive that it wipes out the few performance gains VT gets. Which means my expectation is that Xen/VT/Windows will be significantly slower - there is no theoretical difference between Xen/VT and VMware ESX/VT, since they do about the same thing. I have seen no Xen/VT/Windows benchmarks at all, and I am eagerly awaiting them. I expect Xen/VT/Windows to be, at best, equal to ESX, depending on how good the Xen/Windows paravirtualized drivers are.
• VMware EULAs now allow benchmarking (as of a month or so ago), especially of ESX 3.0. I realize there hasn't been time for anyone to put together good benchmarks - and if I could take the time away from my job, I'd do benchmarks myself, because I believe in being honest about this stuff! But personally, I'm really sick of seeing the study that compares Xen's work against a 5-year-old version of VMware's lowest-end product. And I'm hoping we finally get some interesting benchmarks soon! (Yes, I expect Xen to be faster than ESX, because paravirtualization is theoretically faster - by just a few percent.)
• Paravirtualization ... I'll just point to an event at the Ottawa Linux Symposium this week; Zach is a VMware employee and this paper discusses performance of paravirtualization on ESX.

Two thoughts:

- In terms of summer events, the Ottawa Linux Symposium is supposed to be a great event. I haven't made it to one yet but I've wanted to for a few years. It's July 19-22 in Ottawa (Ontario, Canada).

- [Shameless Plug] This doesn't quite match your description but I thought I'd mention it anyways: I teach at Seneca College, Toronto, and we have a 10-month intensive Linux Administration graduate certificate program that I think is one of the best Linux training programs available. We've had students from all different backgrounds including current staff from large systems vendors. We also throw a great Free Software and Open Source Symposium in October; this year we have Mike Shaver and Neil Deacon (Mozilla), Nat Freidman (Ximian/Novell), Chris Blizzard (One Laptop Per Child), and a raft of others.

And I agree that there's no substitute for getting dirt under your fingernails and actually working with the technology! :-)

Firstly, I apologize for my English (I'm doing my best).

I perfectly agree with some of you: this article is a slashvertisment! The main reason for that is that I previously tried to submit something more descriptive, but it was rejected. That's why I tried again with a slightly different style.

This tool (PTT) inserts trace points into the NPTL to help you to analyze multithreaded applications behaviour. He's not designed for beginners, but for people facing complex multithreaded issues. I also agree with some of you: you can use Java or some others high level languages for programming. But some applications require performance and have to be written in C. That's why PTT can be useful for some developers.

Regards...

• Enhancements to Linux I/O Scheduling

• RapidIO for Linux

• Enhancements to Linux I/O Scheduling

• RapidIO for Linux

• Enhancements to Linux I/O Scheduling

• RapidIO for Linux

Have you even taken a look at the previous year's proceedings? They're all avaliable on the OLS's web site:
http://www.linuxsymposium.org/2006/proceedings.php
All of the talks are highly technincal and most topics covered wouldn't directly affect the average end user. These aren't opinion papers!

Linux does have a "comparable" feature (soon to be merged in mainline) called "kprobes", or "systemtap" (systemtap uses kprobes)

You can see a fairly detailed analisis in the 2005 Proceedings, Volume 2, page 57 of the linux symposium

Also some doc from IBM: http://www-128.ibm.com/developerworks/linux/librar y/l-kprobes.html

also there's a "linux trace toolkit". A post about LTT vs dtrace...whatever, too much flamewar for my taste.

For those of you interested in attending the next one, or who would just like to get news pertaining to the event please consider joining the mailing list.

You can sign up here:
http://lists.linuxsymposium.org/mailman/listinfo/o ls-announce/

You can download the conference proceedings in two parts:

http://www.linuxsymposium.org/2005/linuxsymposium_ procv1.pdf
http://www.linuxsymposium.org/2005/linuxsymposium_ procv2.pdf

You can download the conference proceedings in two parts:

http://www.linuxsymposium.org/2005/linuxsymposium_ procv1.pdf
http://www.linuxsymposium.org/2005/linuxsymposium_ procv2.pdf

• The Canadian version of this law will be enforced in the same way, and
  
• whether this might cause the organizers of these events to change the venue in future years.

A couple of years ago at OLS they had a presentation by someone from Google on their use of Linux in house (this one I think).

At the end of the presentation, he brought up a web page that showed a scrolling list of google queries that were being done in near-real-time. I think it was showing every nth query (every query would have been scrolling too fast to see). He said that they had two versions of this web page - a filtered and an unfiltered one, and he was running unfiltered.

Was it ever funny - people entering some of the craziest X-rated search phrases you could imagine. If he could have left it up I would have stayed there and watched for hours :-)

A couple of years ago at OLS they had a presentation by someone from Google on their use of Linux in house (this one I think).

At the end of the presentation, he brought up a web page that showed a scrolling list of google queries that were being done in near-real-time. I think it was showing every nth query (every query would have been scrolling too fast to see). He said that they had two versions of this web page - a filtered and an unfiltered one, and he was running unfiltered.

Was it ever funny - people entering some of the craziest X-rated search phrases you could imagine. If he could have left it up I would have stayed there and watched for hours :-)

I've written and edited the first book, over 400 pages, and now have started in on book two. I've queried a dozen literary agents who specialize in fantasy fiction, but I've yet to find one who is willing to even read a sample. They all sent back rejection notes that were remarkably similar: Too busy, best of luck with someone else.

Here's someone else you might try. Jim Munroe spoke at this year's OLS about independent media, Linux and free software, and self-publishing. Very interesting, and maybe helpful for what you're trying to do.

...with the exception of the "write the serial modem code" argument, which would be reasonable if the code hadn't already been written for Linux (including emulators), none of your arguments hold water.

You can write and run your software on an ordinary Linux desktop, without even the need for a simulator module for the most part. The same Linux which runs your desktop can also run your 'plane, and for a considerably smaller performance and resource hit than XP (although in real life you'd go to the trouble of having a highly modular kernel and not load very much for the instance on the 'plane).

I have no idea why you'd bother with Atmel chips.

Linux runs natively on the Crusoe, another performance gain.

So you used MS-FS for testing algorithms? Then FlightGear would have given you even more control and oversight over what you were testing.

Reading between the lines, you didn't even look.

• A cache model in which there is one cache for instructions and another cache for data is called a "Harvard-style" cache. This is the model assumed by the PowerPC Architecture, e.g., in the descriptions of the Cache Management instructions in Section 3.2, "Cache Management Instructions" on page 16. Alternative cache models may be implemented (e.g., a "combined cache" model, in which a single cache is used for both instructions and data, or a model in which there are several levels of caches), but they support the programming model implied by a Harvard-style cache. The processor is not required to maintain copies of storage locations in the instruction cache consistent with modifications to those storage locations (e.g., modifications caused by Store instructions).

That's the point here. The PowerPC does not guarantee that self-modifying code will work. There's a long discussion of how to do self-modifying code in section 1.8.1 of the architecture manual. This is mainly for debugger support.

Given this instruction/data separation, it's not surprising that the PowerPC architecture defines a no-execute bit (it's bit 61 of each page table entry). Unfortunately, not all PowerPC models implement it.

These explained a little:

udev presentation (PDF), Google HTML version.

Detailed paper on udev (PDF), Google HTML version.

devfs works fine for me, but since some people (see second link) want thousands of disks I guess I'm not the target market. I mean ... thousands?

As of 2.6.0-test5 or 6, devfs is considered obsolete/deprecated, and AFAIK, will be totally gone from the 2.7 kernel.

The replacement for devfs will be a userspace implementation called udev.

There's an actual paper going around by the guy who came up with an implementation of udev, but I can't think of the URL for it offhand, so here's a little article/discussion on it.

Actually, found the paper. Here's a link to it, in PDF format.

Hans - How is scalability? SGI seems to think that ReiserFS doesn't scale at all with multiple CPUs, unlike ext2. At least according to the paper they presented at the Ottawa linux symposium last month:

(see page 9 of this PDF for the graph)

The implication is a lack of fine-grain locking. Does this new all-atomic, all-the-time implementation automagically bring better locking too?

You have already contributed and are continuing to contribute.

O! Canada. Thank you.

The next generation of Windows is suppposed to include a GPU accelerated compositor similar to OS X's. As far as I know, that's the only one though.

Of interest may also be Xr/Xc. It's not accelrated, but it does have functionality similar to Quartz (Display Postscript). Presumably it *could* be accelerated, so people say. Sorry I couldn't find a better link. There are actual spec docs floating around somewhere on the web.

Oh, did you know that OS X uses runtime statistics to predict what program you will use next and actually preload the binary?? Another one of it's many little tweaks. ;)

OS X is just sweet. :)

Caldera Employee Was Key Linux Kernel Contributor

Christoph Hellwig has been, according to this web page, "in the top-ten list of commits to both the Linux 2.4 and Linux 2.5 tree". The page also mentions another fascinating piece of news, that he worked for Caldera for at least part of the time he was making those kernel contributions:

"After a number of smaller network administration and programming contracts he worked for Caldera's German development subsidiary on various kernel and userlevel aspects of the OpenLinux distribution."

In 2002, he offered a paper on "Linux-ABI: Support for Non-native Applications" which is described like this:

"The Linux-ABI project is a modification to the Linux 2.4 kernel that allows Linux to support binaries compiled for non-Linux operating systems such as SCO OpenServer or Sun Solaris."

Back in 2002, he was described, in connection with his appearance at the Ottawa 2002 Linux Symposium, like this:

"Christoph Hellwig
"Reverse engineering an advanced filesystem
"Christoph Hellwig is employed by Caldera, working on the Linux-ABI binary emulation modules. In his spare time he cares for other parts of the kernel, often involving filesystem-related activities."

So, in short, he was contributing to the kernel and working for Caldera on Linux/UNIX integration at the same time. His work for Caldera was on the Linux kernel ("he worked for Caldera's German development subsidiary on various kernel and userlevel aspects of the OpenLinux distribution"), and he also did work on his own on the kernel. Did Caldera know about his freelance contributions, in addition to knowing about his work for them? What do you think? He used his hch at caldera.de email address when doing it. All contributions to the kernel are publicly available anyway. They certainly could have known. As for his job, his signature on his emails back in 2001 was:

"Christoph Hellwig
Kernel Engineer Unix/Linux Integration
Caldera Deutschland GmbH".

He used the email address hch at bsdonline.org sometimes too, and here you can see some of his Linux-abi contributions. Here are some of his contributions to JFS, Journaled File System. Yes, that JFS. Here he is credited as sysvfs maintainer, and he confirms it in this email, writing, "I've run native sysvfs tools under linux, but as now that I'm Linux sysvfs maintainer I'm looking into implementing free versions of it."

Here is a list of the operating systems that use or can handle the file system sysvfs:

"sysvfs: UNIX System V; SCO, Xenix, Coherent e21
"operating systems that can handle sysvfs: FreeBSD (rw), LINUX (R), SCO (NRWF)"

Here's a page listing by author (alphabetically by first name), with his emails to linux-kernel in June 2003, so he is still contributing.

Here he is listed on the Change log for patch v2.4.17. Here he tells Andrew Morton in 2002 that he will

Why isn't devfs the default now - it's been working fine for ages - for me anyway.

Because devfs is exploitable, slow, and is being ditched by all of the Linux distribution manufacturers. As one former coworker of mine put it so well:

"Devfs is an over-engineered solution to a non-existant problem..."

Seriously though, you need to look at the new work going on, udev, a userspace implementation of devfs.

• MP3 and OGG Audio Files of 2002 OSDN/Usenix Kernel Summit
  
• Identity, DRM and Bad Ass MP3's for link for Digital Identity World's 2002 Archive
  
• Lawrence Lessig, at the 2002 O.Reilly Open Source Conference This presentation on was very widely linked when it was released.
  
• Deploying Network Quality of Service Features presentation which I found useful from the invaluable TechNetCast archives.

Cringely actualy knows who wrote this code Paul McKenney at Sequent (now IBM). And how the code got there (Paul copied and pasted some of his own code into the Linux kernel). And exactly why SCO is wrong (They don't own the 'general concept' of RCU and other tech thought up by Sequent).

This should be what slashdot linked too, not RMS's rant. Cringely did some real reporting and answered a lot of important questions.

Has anyone, besides SCO, looked at the Linux code and tried to determine what might have come from SCO, and what might have come from a common predecessor?

So far four components of the Linux source have been implicated: SMP, RCU, NUMA, and JFS.

I have done a little digging into the NUMA code. IBM has contributed several people who have participlated in developing NUMA under linux. Some names I've run across: Martin Bligh, Matthew Dobson, Patricia Gaughen, John Stultz, Michael Hohnbaum. IBM even has a Linux NUMA news archive. It appears that IBM jumpstarted it's NUMA efforts when it purchased Sequent which was intitally intended to boost its participation in Project Monterey, which is no doubt the origin of SCO's objections.

The most obvious source file for NUMA is /usr/src/linux/mm/numa.c in the 2.4 series kernels. This file contains a comment header stating it was "Written by Kanoj Sarcar, SGI, Aug 1999". This file has been removed from later 2.5 kernels (its gone by at least 2.5.46), appearently because Linux accepted an IBM NUMA patch as reported here. This patch was announced by Martin Bligh and is likely the code in question in this lawsuit.

Has anyone, besides SCO, looked at the Linux code and tried to determine what might have come from SCO, and what might have come from a common predecessor?

So far four components of the Linux source have been implicated: SMP, RCU, NUMA, and JFS.

I have done a little digging into the NUMA code. IBM has contributed several people who have participlated in developing NUMA under linux. Some names I've run across: Martin Bligh, Matthew Dobson, Patricia Gaughen, John Stultz, Michael Hohnbaum. IBM even has a Linux NUMA news archive. It appears that IBM jumpstarted it's NUMA efforts when it purchased Sequent which was intitally intended to boost its participation in Project Monterey, which is no doubt the origin of SCO's objections.

The most obvious source file for NUMA is /usr/src/linux/mm/numa.c in the 2.4 series kernels. This file contains a comment header stating it was "Written by Kanoj Sarcar, SGI, Aug 1999". This file has been removed from later 2.5 kernels (its gone by at least 2.5.46), appearently because Linux accepted an IBM NUMA patch as reported here. This patch was announced by Martin Bligh and is likely the code in question in this lawsuit.

Hey... it looks to me like he works or worked for IBM. Check here. Does this represent any further issues?

This is a very useful application of IPsec. The wavesec project is an example of using IPsec to secure the link between a client and the wireless access point.

This was in-practice last year at OLS where the FreeS/WAN folks set up a wavesec encrypted link, while the folks that were not using wavesec had their traffic snooped and displayed on a monitor.

The problem with using IPsec as a replacement for WEP, however, is that IPsec is higher up on the OSI layer diagram, so more information is left unencrypted than when using WEP (yes, I'm aware that WEP is weak and in this case, won't make a difference, I'm just illustrating a point.)

I understand that the way around that one is to use large pages, to decrease the overhead for each page. Pages can be of variable sizes. I'm not a kernel expert, but there was a talk on this topic at the last Ottawa Linux Symposium.

You must have really had your nose in the books. Ottawa is also home to OLS,
OSW,
the GCC Summit, and the Kernel Summit (see http://lwn.net/Articles/3467/ ). Other stuff too. :)

Check out how this was done at years past at the Ottawa Linux Symposium.

Perhaps this is a bit offtopic, but Debconf 2002 was also announced today. Will holding it in Canada make a difference crypto-wise? Probably not, but it should be a rockin' good time for participants anyway.

It's also been conveniently scheduled to coincide nicely with the Ottawa Linux Symposium. Other than that, more info will be forthcoming within the next couple of weeks.

Apple's always made nice hardware, and geeks appreciate that. The only reason why I bought x86 is because of MacOS. But, Linux on PPC means I can now run my favorite OS on kickass hardware. It's win-win.

I run YellowDog Linux on my Titanium G4 notebook. Why? Performance is *excellent*, plus the notebook ain't to shabby to look at. Extra wide screen, good hardware integration. Works for me. In fact, every year I've attended the Ottawa Linux Symposium, I've seen more and more people lugging Apple hardware around, running Linux. We're not talking your average joe users either, but serious developer types. I seem to recall a few of the Samba fellows typing away on Powerbooks during the keynote...

MOL (Mac On Linux) is a nice tool for those that do run Linux on their Mac, yet occasionally need to boot into MacOS. It saves a reboot, and can be quite handy when you need to playback a Quicktime file, or something along those lines. I'd rather have a native Linux player of course, but since Sorenson won't disclose the codec, MOL allows me to run Linux yet still access one or two of those quirky Mac apps. ;)

As another poster mentioned, it's like VMWare for the Mac. If you can postulate a use for VMWare, then it shouldn't be too difficult to figure out why some people like MOL.

Hopefully a startup might be able to manage the distribution better than Corel has, with tighter focus and better communication with the community. The downside is that I will miss Corel's excellent presentations at OLS.

It is really telling that the Creation company, which runs for-profit Science Fiction and Star Trek Conventions is involved this these people. I remember my first and only Creation Con back in the Early '80s...back then Fan run cons where charging $20 for a weekend membership, and it cost me $25 for one day at the Creation Con. Creation Cons are consistantly over priced, have crappy programming and driven up the speaker's fee s for guests so much that the average convention cannot afford guests. Thankfully, SF Book oriented conventions never really became a target for Creation...we still have good, inexpensive conventions for people who enjoy Science Fiction and Fantasy Books!
ttyl
Farrell J. McGovern
Founder, CAN-CON Ottawa, Canada (That is the same places a OLS)