linux.conf.au 2008

By Jonathan Corbet
February 6, 2008

linux.conf.au has an interesting structure which differentiates it from most other events. Every year, a completely new set of organizers takes over the event, moves it to a new city, and puts its own stamp on it. They have a great deal of freedom in how they run LCA, but there is still a group of Linux Australia members and past organizers who keep an eye on things and help ensure that the event does not run into problems. The result is a conference which has a lot of fresh energy every year, but which is also reliably interesting. Many attendees consider it to be one of the best Linux events to be found anywhere in the world.

This year, LCA was held in Melbourne, Australia; the organizing team was led by Donna Benjamin. The now-familiar LCA formula was followed, but with some small changes. The tutorial day is no more, replaced by relatively short tutorial sessions on each day. The traditional auction for charity was also gone this year; instead, a raffle (with Greg Kroah-Hartman's 2.6.22 contributor poster as the main prize) yielded some $1000 for a local penguin refuge. The raffle was certainly a lower-pressure, less alcohol-fueled way of raising money, but LCA without Rusty Russell as auctioneer just isn't quite the same. That quibble notwithstanding, LCA 2008 was an interesting, well-organized, and well-attended event. Ms. Benjamin and company have certainly upheld the standards for this conference.

A number of LCA talks have been covered in separate LWN articles, and a few more may yet follow. This article will quickly review a few other high points, as seen from your editor's perspective. It's worth noting that videos for almost all of the talks have been posted on the conference web site.

Certainly one high point came on January 30, the day that LWN celebrated its tenth anniversary. The crowd sang a rousing - if not entirely harmonious - version of "happy birthday" after Bruce Schneier's keynote. The following morning tea featured special LWN muffins; they were, much to your editor's delight, of the intense chocolate variety. It is hard to imagine a better place or time to celebrate to celebrate ten years of LWN.

While most LCA presentations are quite technical in nature, there are exceptions. Australian lawyer Kimberlee Weatherall's talk on legal issues was called "Stop in the name of law"; it covered a number of topics of interest to a global audience. Kimberlee, it's worth noting, was the recipient of the "Rusty Wrench" award for service to the free software community at last year's LCA in Sydney.

The Digital Millennium Copyright Act, she noted, is ten years old now. At this point, the debate on its anti-circumvention provisions is essentially done, and anti-circumvention has won; she is not expecting to see any major changes in countries which have adopted such laws. The music industry may be moving away from use of DRM, but "they were never very good at it anyway." DRM is still going strong in other areas, such as movies and subscription television.

Similarly, the fight to end software patents is over, and we have lost. There are incredible numbers of software patents issued every year; every one of those patents represents a significant investment by its owner. The total amount of investment in these patents is huge; that amount of money is almost impossible to displace. It is also very hard to define what a software patent really is; there are thousands of them in Europe, which ostensibly does not allow software patents. No matter how the rules are written, lawyers will find a way around them.

What is happening on the patent front, instead, is a more constructive engagement with the process. Some reform is happening in the US, as a result of the KSR decision and various attempts to mitigate the costs associated with patents. So the situation might improve slowly over time.

GPLv3 is out. It now has to pass two tests: the market test (will projects use it?) and any legal tests which might be brought. Kimberlee expressed some doubts on whether GPLv3 will really hold up in court, but did not elaborate on them.

There is a new threat out there which we should not underestimate: the push to force copyright enforcement duties onto ISPs. This effort takes two forms: getting "infringers" disconnected, and requiring ISPs to filter data passing through their networks. There are a lot of problems with either approach, but that is not stopping the industry (and others, such as anti-porn crusaders) from pushing hard for ISP responsibility. This is a fight to watch.

So what should the free software community do? Not much, says Kimberlee, except to keep coding. The production of good code brings us allies with money, and that's what we're going to need. As long as we are successful, people will go out of our way to protect us. Keep doing what we do, and things should come out OK.

Anthony Baxter is the Python release manager; he was also the keynote speaker for the third day of the conference. He is, to say the least, an entertaining speaker, so this would be a good one to watch on video. The talk was about coming changes in Python, and Python 3.0 in particular. The 3.0 release, he says, is "the one where we break all of your code." It's the first backward-incompatible update of the language (at least, if you don't deal in C extension modules).

There are a lot of changes to the language which your editor will not repeat here; they are well documented on the Python web sites. As noted, many of these changes will cause existing code to break. This is being done, says Anthony, because the Python language is now 16 years old. Like all 16-year-olds, it has a number of annoying features. It's time to clean out a lot of accumulated cruft and get back to the minimal, "there is one way to do it" vision that has always driven the language.

Perhaps what's most interesting is what won't be done. The language will not be bloated - it will stay Python. There will be no braces; white space will still be used to mark blocks of code. The much-criticized global interpreter lock will remain. And, importantly, this will be an incremental (if big) update - there will be no overall rewrite of the interpreter. The experience of certain other projects (being Perl 6 and Mozilla) shows that total rewrites tend to be much longer, more painful affairs than anybody might envision at the outset.

There will be migration tools, of course, and warnings built into the forthcoming 2.6 release which will point out things that may cause migration difficulties. The 2.x series will be supported for some years into the future. And, says Anthony, there will be no Python 4.0 release. This is their one chance to break everything and start over, and they plan to get it right this time.

Dave Jones is the head maintainer for the Fedora kernel. At LCA 2008 he took a break from pointing out user-space problems and talked about "a day in the life of a distribution kernel maintainer." The real subject of the talk was the process that the Fedora project goes through to put together the kernels they ship.

There are currently three developers working on the Fedora kernel (Dave, Chuck Ebbert, and Kyle McMartin), and "several dozen" working on the RHEL kernels. Most of the RHEL folks are doing backports of fixes, drivers, etc. to the older kernels used by RHEL releases.

Once a kernel has been chosen for release, it's time to start adding patches. Some interesting numbers were put up at this point. Red Hat Linux 7 had 70 patches added to its 2.2.24 kernel. That number went slowly up, to the point where Fedora Core 6 had 191 patches. There are currently 63 patches added to the Fedora 8 kernel, though that may grow over the life of this release. By comparison, RHEL 5 is shipping a 2.6.18 kernel with 1628 patches added to it - a very different world.

There's all kinds of patches which go into a distributor kernel. These include security technologies (ExecShield) which have not made it into the mainline, changes to some default parameters, the silencing of certain "scary messages" which tend to provoke lots of needless bug reports, out-of-tree drivers, patches which help debug problems found in the field, stuff which has been vetoed upstream, and more. Then it's a matter of putting the package and dealing with the subsequent bug reports - lots of them.

The closing ceremony included the traditional introduction of the organizer for next year's event. This event will go, for the first time ever, to Hobart, Tasmania; see MarchSouth.org for more information. There is some information on what this team is planning in the bid document [1.6MB PDF]; your editor is intrigued by the following: "The official Speakers' Dinner will be held at a mystery location south of Hobart following a 40 minute river cruise on a high speed luxury catamaran." It's never too soon to get that talk proposal together.

Finally, the last few LCA events have included the passing of the "Rusty Wrench" award to somebody who has performed a great service to the community. Recipients so far are Rusty Russell (after whom the award is named), Pia Waugh, and Kimberlee Weatherall. The Rusty Wrench was not awarded at LCA2008, though. It seems that, in the future, the Rusty Wrench will be part of an extensive set of awards which will be handed out at a separate "gala dinner" event held in the (Australian) winter. The awarding of the Rusty Wrench was a nice LCA feature which will be missed, but, then, there are advantages to having another excuse to visit Australia.

Index entries for this article
Conference	linux.conf.au/2008