Joshua Wulf

Hi there, in the blog on Sept. 16 you said:

"First off, it seems Fedora developers did not define the UTS_RELEASE in the kernel version header."

This is something that has changed upstream in the kernel:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=206158
"UTS_RELEASE has moved to linux/utsrelease.h"

You might like to update your script to take this into account.

regards,
--josh

Funny thing is - I blogged this very change on July 19th. Too much information to keep track of... Thanks to the anonymous commenter for the heads up.

Yesterday was the third (and final) test release candidate for Fedora Core 6. Initial support for Fedora Core 6 was brought in the packaging scripts with the 8.28.8 drivers. With the release of FC6T3, it seems the fglrx drivers are working but it does (currently) require a few tweaks on the end of the user.

First off, it seems Fedora developers did not define the UTS_RELEASE in the kernel version header. When running --buildpkg Fedora/FC6 on a clean FC6T3 install, the build will error out. The error, however, resides with the kernel. Within /lib/modules/2.6.17-1.2630.fc6/build/include/linux/version.h the UTS_RELEASE is not specified. Adding #define UTS_RELEASE "2.6.17-1.2630.fc6" to version.h will correct this problem so the RPMs generate fine.

From ATI Linux Blog.

Not that we're into making it easy for people to install proprietary software on their all-free Fedora system, but we're not into making it hard, or pushing out "broken" software.

I'm running 2647 and the UTS_RELEASE is still not in there. Should it be?

I found Democracy TV while searching for a video podcatcher to receive and watch Chaz's Yogamazing video podcasts in.

Here is how I got Democracy TV working on my Fedora rawhide system

First of all I downloaded the source package for Fedora Core 5 from http://www.getdemocracy.com/downloads/fc5src.php

The rpm package available from the Democracy TV site would not install due to failed dependencies, mozilla-devel prominent among them.

The build of the source (rpmbuild -ba /usr/src/redhat/SPECS/Democracy.spec) failed with a similar error. The mozilla-devel package does not seem to be available for Fedora rawhide. I found a mozilla-devel rpm for FC 5 here: http://linuxsoft.cern.ch/fedora/linux/core/5/i386/os/repodata/repoview/mozilla-devel-37-1.7.12-5.html

and force installed it using rpm -Uvh --nodeps, as the mozilla installed on my rawhide system is 1.7.13.

After this it still wouldn't build, and then I found this thread: http://www.fedora-france.org/modules/newbb/viewtopic.php?topic_id=12245, in French.

Following the instructions there, notably:

as root :

cd /usr/lib/mozilla-1.7.12/
ln -s /usr/lib/mozilla-1.7.13/libgtkembedmoz.so .
ln -s /usr/lib/mozilla-1.7.13/libxpcom.so .

I was able to convince the package to build, and then install it. Installing it required me to get gnome-python2-gtkmozembed, which luckily for me this time, was available in the rawhide repositories.

Then it appeared in the Sound & Video menu.

Now that I have it, I don't know if I can actually watch the podcasts with it. Chaz provides his podcasts through iTunes. Still trying to figure that out...

There was a little bit of confusion about Tim Lauridsen's depcheck plugin, so allow me to elaborate.

Depcheck will remove any packages from a yum operation whose dependencies cannot be satisfied, and continue processing the ones whose dependencies can be satisfied. No, depmod does not allow you to do a yum install with the equivalent of rpm --nodeps, but wouldn't that be cool?

As Seth explained in his yum tutorial at LCA last year, yum is not a dump truck, it's a pipe, and you can smoke whatever you want in that pipe by using the yum plugin architecture.

Seth said he did this because he got constant requests for "crack", and he wasn't going to put crack into yum, but he wanted to give people the freedom to smoke crack if they want to.

A --force plugin would be some serious crack, but depmod isn't it.

Sorry to disappoint any crackheads out there who were looking forward to trashing their systems...

Yum just got a whole new twist with yum-2.9.5-4. Man, this thing is so fast it could get a speeding ticket.

It's got a C meta data parser, and with Tim Lauridsen's depcheck plugin it can even do an update when some package dependencies can't be satisfied.

For anyone who can remember, the speed of yum has been an issue for me for a while, with Ubuntu toting friends updating their machines in a fraction of the time it took me to install a package using Fedora's updating and package management system (I don't admit to the existence of that iMac running 6.06LTS under my desk).

Anyway, that's history. There's a new yum on the block, and the word is that 2.9.6 will be even more smoking.

Big ups to Seth and Tim for their work, and props to Rahul for the heads up.

This is a momentous day for me.

I got it going, thanks to something that I overheard in #anaconda.

You can find out what to do by visiting this thread on vmware.com.

Just be aware that this problem has arisen due to a change in the kernel which forces code that uses the lock dependency infrastructure to depend on a GPL-only symbol, which b0rks any devil code like VMWare.

Getting around it like this will probably endanger the eternal destination of your soul, but at least you'll have your VMWare goodness before you get all toasty... ;-)

At the moment I'm trying to get VMWare Workstation 5.5 running on Fedora Core 6 Test 1. Right now I'm working on some documentation for RHEL 5 and the build chain is all based on RHEL 4, so I need to have a RHEL 4 environment.

I downloaded VMWare Workstation 5.5 and tried to get it going.

First problem I ran into was the vmware-config.pl complaining that there was no kernel source. It requires the kernel-headers in order to compile a kernel module. I installed kernel-devel for both i586 and i686, but it still wouldn't pick it up, even when I explicitly told it where to look.

When I manually entered the directory /usr/src/kernels/2.6.17-1.2405.fc6-i686/include/ it would complain that the header version (@@VMWARE@@ UTS_RELEASE) did not match my running kernel, and that "even if the module compiled it would not work with the running kernel".

Use this command:

echo /usr/src/kernels/$(uname -r)-$(uname -p)/include

(from here) to figure out what the kernel header directory should be for your running kernel.

OK, so mine was fine, but it still wouldn't work.

A little Google love found me this thread on a Red Hat mailing list.

So I went to http://platan.vc.cvut.cz/ftp/pub/vmware/readme.txt, read through it and picked up the copy of vmware-any-any-update102.tar.gz. I unzipped it and ran runme.pl.

That got me further. Next I got an error advising me to run "make prepare" in the kernel header directory. I did this and it got a little further, then bombed out again telling me:

hostif_compat.h: error: expected ':' or '(' before 'UTS_RELEASE'

I Googled some more, then found this exchange between the kernel Gods.

UTS_RELEASE, it turns out, has been moved from version.h to utsrelease.h in kernel 2.6.18-rc1 and above.

The kernel I'm running on FC6T1 identifies itself as 2.6.17-x, but apparently has UTS_RELEASE in utsrelease.h.

The VMWare hostif_compat.h file has the following in it:

/* Retrieve UTS_VERSION and UTS_MACHINE if there is such information available */
#include <linux/compile.h>
#endif
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 18)
#include <linux/utsrelease.h>
#endif

which obviously doesn't work on my kernel, which exhibits the 2.6.18-rc1 behaviour, but identifies itself to VMWare as 2.6.17.

I untarred /usr/lib/vmware/modules/source/vmmon.tar and edited hostif_compat.h to change it to read:

#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 17)

This enables the module to compile ok, however when it tries to load it at the end of the VMWare configuration I get the following error:

make[1]: Leaving directory `/usr/src/kernels/2.6.17-1.2405.fc6-i686'
cp -f vmnet.ko ./../vmnet.o
make: Leaving directory `/tmp/vmware-config14/vmnet-only'
Unable to make a vmnet module that can be loaded in the running kernel:
insmod: error inserting '/tmp/vmware-config14/vmnet.o': -1 Unknown symbol in module
There is probably a slight difference in the kernel configuration between the 
set of C header files you specified and your running kernel.  You may want to 
rebuild a kernel based on that directory, or specify another directory.

Anyone have any ideas, or already done this? As you can probably tell - I'm stabbing in the dark here...

Still trying to track down the source of the failure of wireless in the ppc rawhide kernels on my 12" ibook G4. (BZ #193082. Come on - anyone else out there seeing this problem? I'm feeling all lonesome here :-)

It's a little inconvenient because in the FC5 kernel wireless works but suspend blanks my screen on resume, requiring a reboot. In the rawhide kernels suspend works a charm, and wireless goes south...

At the moment everything mono seems borked on my system. Tomboy, banshee, f-spot, and beagle all give me gtk-related errors on startup. I lodged a bug #192183 against beagle-search, because that's where I first noticed it, but now I realize it's all the mono apps.

I've got a similar problem with rhythmbox where it blows up with a gtk related error when I try to import files or folders.

Anyway, small inconveniences on the way to a world conquering desktop...

Yet another customer who opens a ticket saying: "System crashed - kernel panic. I need to find out root cause. Please advise".

No sysreport, no information about what they were doing with their system. Nothing.

Hello - is this Red Hat, or Oracle? As in the Oracle of Delphi: "Excuse me one moment while I check my crystal ball...."

Hang on a minute.... that's an idea. So I call Franciso Cruzat, based on a recommendation from "The Net's Best Psychics - Your Guide to a Good Psychic Reading"

10 minutes and AUD$45 later I call that customer back: "Sir, your RAID controller firmware requires upgrading - I think you'll find that you're using version 1.34g - you need to go to 1.36b."

No, it didn't really happen - but it should. We should get psychics working the help desk - the customers seem to think that we already have them. Even if we had to put the psychics through RHCE training before we put them on the phones, we could get customer satisfaction up immensely. Hey, if we can get the really good psychics on board, we could call the customers up, instead of the other way round: "Sir, please make sure that you have a full backup, you're going to experience a drive failure this week".

Now that's technical support.

I must say that Rawhide is looking pretty spanking. The new icon theme totally rulez. Nice crisp feel. I haven't booted into OS X for weeks. The only thing that OS X can do that FC5.89 can't is flash embedded in the browser, and proprietary codecs like WMV 9.

Otherwise we're getting parity, and even coming ahead in some areas of usability and sex appeal.

The thing that really stands out to me the most, and it's what has always impressed me, is the rate of change. If you look at how far behind we've been, and how fast we're catching up, it seems that projecting that curve out further, GNU/Linux has the potential to have technical leadership, even in the desktop space, within a year or two.

At the moment I'm working on isolating the breakage of the bcm43xx driver in rawhide. FC5's kernel works fine for me, but the rawhide kernels don't. I'm running all rawhide except for the kernel.

Since an update or two ago programs have started randomly quitting on me. There are all internet enabled ones, but I don't know if that's significant, because most of what I do is interacting with the web. I started launching them from a terminal to see if I could get more information, and this is the message they all give:

Gdk-ERROR **: The program 'yumexmain.pyc' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadRequest (invalid request code or no such operation)'.
(Details: serial 9353718 error_code 1 request_code 0 minor_code 0)
(Note to programmers normally, X errors are reported asynchronously;
that is, you will receive the error a while after causing it.
To debug your program, run it with the --sync command line
option to change this behavior. You can then get a meaningful
backtrace from your debugger if you break on the gdk_x_error() function.)
aborting...
/usr/share/yumex/yumex: line 5: 17653 Trace/breakpoint trap /usr/bin/python /usr/share/yumex/yumexmain.pyc $*

Anyone know anything about this?

In other news, I got my RHCE a month or so ago, opened a yoga center, Atma Yoga, and I've started a yoga teachers training course in my spare time, which I don't have so much of right now, for some strange reason....

Oh yeah, MC Hawking wrote me to say that he's not dead, and that he'll come on the next Fedora Reloaded podcast to prove it.

Fedora Reloaded 5 podcast is out now. It's a special tribute to the late MC Hawking where we look at "The Future of Fedora".

We talk with Guildford, UK's David Woodhouse about the future of Fedora on PPC, chat with Max Spevack, the newly appointed Fedora Project Leader, get Greg DeK to speak out on behalf of the people (which isn't hard to do), and talk shizz, all to the accompaniment of classic phat tracks from the Hawkman.

Hyperposting will stop tomorrow, the last day of LCA. On Sunday morning I'm on a 7.20 am flight to Christchurch, and thence to Taupo, and I'll be off-net for a week.

Tonight was the Linux Conf AU dinner at the Dunedin town hall. At the end there was an auction to raise money to donate to the John Lions chair that is to be established at the University of New South Wales ( which is incidentally where LCA 7 will be held next year).

Rusty Russell was the auctioneer. Now I didn't get to hear Rusty speak at this conference, but I definitely knew who he was. Back in the office in Brisbane we were discussing Australia's disproportionate contribution to open source globally and we were running through the hometown heroes. mkearey, who like a few others at Red Hat used to do networking in university dorms, and is my personal recourse for all things iptables related, gave props to Rusty Russell.

Rusty was amazing, a master of theatre - even managing a Shakespeare quote at one point of the show. The item being auctioned was a copy of John Lions' Guide to version 6 of Unix, with source code, signed by Linus Torvalds, Ken Thompson, Dennis Ritchie, and a number of other open source superstars. The auction went for an hour, and by the end of it a cartel had bid $10,000 for the book, as well as:

• Jeff Waugh shaving off his hair


• A mention for whoever they nominate (Miguel de Icaza? <- joke!) in the About dialog of KDE 4


• Naming rights on the upcoming Planet 1.0 release


• Dave Miller shaving off his beard and moustache


• Russell Coker growing a beard for the next LCA


• Groggy Greg shaving off his beard


• A mention in the release notes of MySQL 5.1


• 8 bug fixes of their choice in KDE

Linux Australia is matching the $10,000 from the auction, and Usenix is matching that total - for a grand total of $40,000 raised toward the UNSW John Lions Chair.

Benno got up and explained that the cartel was formed of ex-UNSW students who studied in the culture established by John Lions in his 23 years as a professor there, and that they felt so strongly about it that they wanted to contribute back to that culture.

I left with a lot of the LCA massive just before midnight. The keynote tomorrow morning is Mark Shuttleworth speaking on "How to Improve Collaboration Between Open Source Projects". Mercifully it doesn't start until 10 am. I'm getting pretty worn out. I don't drink alcohol, and I can't understand how other people are managing with the late nights and the intense, action-packed days, and drinking at night. Amazing.

The guys running this event must be rushed off their feet. There is so much on. And it's just not possible to catch all of it. I missed Ajax's talk on X memory management, and Carl Worth's Cairo tutorial.

We did the Red Hat BOF tonight. Dave Woodhouse, Dave Howell, myself and Seth Vidal (disclaimer: who doesn't work for Red Hat!), and Menno Smits were there.

Before we went I got the low down on the Fedora distro upgrade situation from Jef Spaleta and Seth Vidal.

If you go here you can see that Fedora allows an anaconda upgrade of one version of Fedora to the next. Check out Jef's comments to find out about the current situation of this in FC5 testing.

Doing an upgrade of a live system without rebooting, which is actually what our friend was after, is not officially supported, or recommended. Seth explained that nasty things can happen if you try this. However, Seth himself does it, and has a page where he talks about it - here.

There were about fifteen people, and I introduced them to Seth, who then kept them entertained for about an hour, at which time I gave everyone a t-shirt and handed out some FC4 DVDs for the peeps who wanted them.

A number of the people who came are in an academic environment. There are a number of options for academic environments. I believe there is a special academic pricing for Red Hat Enterprise Linux (I'll have to check on the details of that - I'm not in sales). Seth uses CentOS (he's a maintainer) at Duke, which is a rebuild of the Red Hat Enterprise Linux source packages. It costs $50 a year for a machine. You don't get technical support with that, or the centralized management of your machines that you get with Red Hat Network or Red Hat Satellite (all of which can save you a lot of time), and some third-party apps like Oracle that are only certified on RHEL won't run, but if you don't need those things, it can work for you.

If I had a business which had the skillset internally I'd probably run CentOS like Seth does. Otherwise I'd just pay for RHEL. Whatever the yearly subscription is it might seem like a lot for software, but you're not paying for that, you're paying for support staff, and for support staff that's pretty cheap. The guys that I work with are top notch and can fix all kinds of problems. A lot of the time I just help people to administer their machines. Sometimes we work to solve complex problems during implementations (they always ring just before their deadline!). Sometimes I connect people with engineers to get bug fixes done (people with subscriptions get priority on engineering resources). There are a lot of people whose work is something other than Linux, and they don't have time to spend on getting their system working, and they don't have the money to hire a consultant or a full-time Linux person. It's perfect for them.

The other night I was explaining the Red Hat business model to someone on the bus. The original idea? Give the Linux away for free, and make money selling the t-shirts. It actually hasn't changed much. We still give the software away for free. You can get it in binary form from CentOS, White box or Tao, or build it from source yourself, like they do. We make more money supporting it these days than from selling t-shirts, but t-shirts are still in the mix.

Have to get some more t-shirts. I gave away all the medium sized ones, and now there are only a few XL ones left. The new Fedora logo will look good on a t-shirt.

Last night on the way out to Larnach Castle one guy on the bus was, I kid you not, recording the waypoints on his GPS. Now when the guy next to you is recording the waypoints on his GPS you are either at a geek conference, or travelling with Mossad agents.

We started chatting, and he filled me in on the state of GPS, including DGPS, or distributed GPS. He also told me about another talk that I missed, about the fastest Debian computer in the world. I had read that title during the Debian mini-conf stream, but was in something else at the same time. There are so many great talks going on that it's impossible to catch all of them.

"The fastest Debian computer in the world" was a talk given by Keith Packard about his hobby of rocketry. I tried my hand in solid fuel rocket building while I was in high school, but this is of a whole different magnitude. From what I was told the rocket goes up to 75,000 feet, and it has two GPS units in it, one in the nose and one in the tail, and with these they can compute the bearing of the rocket as well as its position.

Anyway, now completely blown out at having missed all of Keith Packard's talks I scanned my timetable today and was instantly on a talk entitled: "A modest proposal to help speed up and scale the linux networking stack", by a guy called Van Jacobson. This one jumped out because it looked like another "Mouse Pointer Redirection in X". The title was completely non-descript, so it had to be the shizz.

I went along, and this time I scored! Apparently (as everyone else no doubt already knows) Van Jacobson is the man in networking - from the old, old school.

Like Keith Packard he was a very non-imposing personality, starting off in a soft voice that we had to strain to hear. I had to run out part way through to secure my flight for Sunday, but I caught the first half, and I can now say that I've seen Van Jacobson. Steve Hanley has the lowdown on the talk.

Tonight is the Red Hat / Fedora Birds of a Feather session in Castle One. It's at 6 pm. See you there.

Damian Conway gave the keynote this morning here at Linux Conf AU 2006, and it was awesome. Damian Conway is one of the Perl 6 Design advisors. He is a professional speaker who makes his living addressing groups of people on technical subjects.

Today's talk should be on Presentation Zen man. It was great. Awesome humour - complex jokes involving verbal and visual gags, lots of references that would have meant nothing to a non-geek audience. Great momentum, a large number of slides, fast transitions and tight integration with his speech. At the Presentation Aikido mini-conf, if memory serves me correctly, Damian advised 10 hours of preparation for every hour of speaking, and said that he personally does 20 hours of prep work. He spoke for an hour and a half and landed exactly on time at 10:30 am - that would mean at least 30 hours of preparation, but I wouldn't be surprised to hear that it took more. To prepare the slides alone would be a feat.

My mind is blown by that presentation, and not just that one. There seem to be so many people here who are not only technical geniuses, but also social geniuses. Damian's talk was on "Technical and Social Lessons from the Perl 6 Development Process".
He spoke about technical design considerations and also social considerations for open source project work. Amazing stuff.

Last night we went to Larnach Castle, New Zealand's only castle, for the Professional Delegates Networking Session. I met Manish, one of the developers of the Flock web browser, which is essentially Firefox with Bling. We did an interview this morning about Flock for a podcast (either Linux Australia or Fedora Reloaded, or both).

The guy who built Larnach Castle (summarizing here from the history room there) had three wives successively. The first one died, and he introduced a Private Member's Bill to Parliament (he was an MP) to have the law changed in order to be able to marry her half-sister, against the objections of his children. He quarrelled with his children over their inheritance and gipped them out of it. She died, and in his 50s he married a 30-something year old woman, who then alledgely had an affair with his son. He shot himself (in the Parliament Building in Wellington), and the son and his wife then battled the other children in court for the inheritance, eventually losing. His son then later shot himself.

Not much of a fairy tale for the castle, is it...

Yesterday was Australia Day, and as the night wore on there were successively louder chants of "Aussie Aussie Aussie! Oi Oi Oi!". I was on the last bus back, and there was a loud chorus of Kiwi hits, including Dave Dobbins' "Slice of Heaven" and Crowded House's "Always Take the Weather (with you)" by some patriotic Kiwis who were feeling a little challenged by the preponderance of Australiania flags and tattoos - including one on the Fitz's forehead. Conrad Parker had a nice one on the side of his neck too.

Yesterday during a break in Seth's talk I spoke with someone who uses Fedora in a commercial environment. He raised an interesting point that warrants some consideration.

Fedora has a release cycle of 6 months and a support cycle of 1 year. It's an aggressive cycle designed to keep the distro on the bleeding edge, with the latest technologies available as soon as possible, and to keep as much resource as possible focused on the point of the spear.

However, for people who want to use Fedora in a commercial setting as a production server, this is too short. Other distributions, which shall remain nameless, have a much longer supported lifetime. Fedora Legacy, the project that supports end-of-lifed Fedora versions, is not quite there yet, according to this gentleman.

The other problem is that there is no officially supported method to upgrade from one version of Fedora to the next one six months later. This gentleman asks: "Can we have a longer supported lifetime, or a means to put in a CD, hit a button, and go to the next version, please?"

I guess if you think of Fedora as a beta test for Red Hat Enterprise Linux then the six months, upgrade or slide into oblivion after one year thing makes perfect sense. "If someone wants a fully supported solution that lasts for 7 years they can just purchase Red Hat Enterprise Linux. As far as supporting Fedora beyond one year, well if there is demand for that then let the wider community provide it."

This may well become the situation in the future. In the meantime, an upgrade path would seem to me to be easier to provide, in order to encourage people to use Fedora.

I know that it can seem like Fedora is competing with RHEL: "Hang on, if people can use Fedora in a commercial setting easily, then they won't buy RHEL. Let's put some obstacles in the way, like not supporting it for more than a year and not making an upgrade path available." I'm not suggesting that there is an explicit conspiracy along these lines, but it can seem like this.

The thing about RHEL is that it is most useful for companies where they have to have a "throat to choke" if something goes wrong. One gentleman I was speaking with on the bus last night related how in his corporate environment they just don't get it when the internal technical staff do the support for the machines. They want to see a support contract with a vendor name on it.

Other corporate environments demand stability over many years. They want the same versions of software with backports of security updates and bug fixes. These guys need RHEL.

Only very small, or very bleeding-edge shops are going to change their version every six months or every year using a version upgrade option. However, these are just the guys that we need to be using Fedora.

A side note here - this gentleman said that a thousand dollars for RHEL was just too much - there is a basic subscription for RHEL ES (basic means you get software updates and can get web-based technical support) for just US$349 per year - so it's not as expensive as it can sometimes seem.

I've never seen it, but there's a reality TV show called "Pimp My Ride", where they take someone's car and rice it up. A reality TV show for LCA could be called "s*X Up My Conference Title". I thought of this because to the ordinary punter it can sometimes be hard to determine which talk is going to have the s*X factor (for them) from the title. The other night at Google Drinks I was chatting with Keith Packard. He gave a talk on Wednesday about "Mouse Pointer Redirection in X", which to me, sounds about as unfascinating as it gets. How sexy can a mouse pointer be? I asked him what it was about and he explained that it was about "wobbly windows" and mouse clicking in them while they're wobbling. "Egads man!" I exclaimed. "Why didn't you say so? I'm all about wobbly windows!"

Keith's talk was on at the same time as several others, obviously. I was sitting earlier in the day with some friends scanning the program to decide what to attend. I overheard someone saying: "Cell processor man, that's sexy." I couldn't really understand the topic, but I had to agree that Cell is sexy - Playstation 3 anyone? Playstation 3 beats "Mouse Pointer in X" any day. However, it doesn't beat wobbly windows, not by a long shot.

Today we were discussing yesterday's events, and I was lamenting having missed the wobbly windows in favor of the Cell processor talk, given by Jimi Xenidis (that's pronounced ex-in-ee-dees) from IBM, based on research work that he and off-sider Amos Waters did. "Man, all they said was: 'We had an idea, and we tried it, and it didn't work.'"

"What they were talking about was using a hardware device to accelerate zeroing memory before release while deallocating it, yada yada yada..." (more detail from someone who got it) Ted and Tim, with whom I was having lunch, and who were also at that talk, started knowledgeably discussing the hardware and the pros and cons of the approach based on the intimate details of the Cell processor, which they had learned from Jimi's presentation. Remember that IBM designed and built the Cell processor, so those guys know how it works. When they try out edge case experiments, they're really pushing the envelope. I had a sudden realization that I hadn't actually understood 5% of what Jimmy was explaining, and I had missed the wobbly windows.

If only the title had mentioned the wobbly windows!

I reckon that Damian Conway, the "Carson of the Conference Keynote Title", the man who gave the "Presentation Aikido" presentation during the mini-confs and is giving a Keynote entitled "Sex *and* Violence: Technical and Social Lessons from the Perl 6 Development Process", would be the indicated man to present the show "s*X Up My Conference Title". (Actually I got the idea from a presentation that Damian does called "Geek Eye for the Suit Guy").