Ballarat-3350 – Jack Scott's Blog

linux.conf.au 2012 – Day 5 (Friday)

This week I’m at linux.conf.au, the southern hemisphere’s premier open-source conference. This year it is being held in Ballarat, about an hour’s travel from Melbourne. I’ll be documenting the trip and conference as much as I can given the limits of my enthusiasm and awakeness.

Friday 20th January:

Friday is the last day of the conference, and everybody is starting to look tired; it’s a full-on week. But, before we all go home, there are just a few more excellent talks to attend. The first of these was Friday’s keynote, given by Jacob Appelbaum, and what an amazing keynote it was. Jacob talked about the state of surveillance states. He explained what they are doing to keep track of all of their citizens, and the special measures that have been put in place in the last few years (mostly since September 11) that significantly curtail our freedoms in the name of privacy and safety. A few choice quotes from the talk:

Free software is for freedom, open source is for business solutions.

Be the trouble you want to see in the world. [It’s in my notes, but I’m pretty sure it was actually just written on his shirt]

90s Nihilism: I have nothing to hide.

The data kept about you in [server] logs around the world tells a story that is not necessarily true, but is made up of facts.

This talk flowed on nicely from Senator Ludlam’s talk at the Penguin dinner.

After morning tea, I watched the talk by Rusty Russell and Matt Evans about why UNIX has been getting bigger over time (in terms of binary bloat). It’s mostly due to new features, but also because of the infrastructure that modern systems have and the libraries that are statically linked in these days (glibc is basically just bloatware). Also in this session I attended the talk by Simon Horman on Open vSwitch. It’s really interesting content, but the presentation was a bit dry. It’s definitely something I want to check out when I get home though, as it could be quite useful for me when I have VMs set up in Linux. The support for VLANs makes it a much better choice than standard Linux network bridges.

During lunchtime there was a meeting between a group of Tasmanian delegates, and it was decided that the Hobart Linux User’s Group should be started up again. So if you’re reading this, like Linux and live in Hobart, get in touch!

After lunch was the best-of sessions. These were talks voted for by the delegates that they wanted to see again, or missed the first time around. I watched two fabulous talks. The first was on Codec2 (presented by David Rowe), an audio speech codec that uses 1400 bits/sec for transmission, which is a 500x improvement on raw 16bit 44.1kHz audio. Very impressive. The second was on the freedom box project (presented by Bdale Garbee, which is a platform for developing easy-to-use home servers oriented towards federated social networking services (such as Status.net or Diaspora). This followed on nicely from Appelbaum’s talk that morning, giving a solution to some of the problems that were outlined.

The final session of the conference was the lightning talks. The real highlight was watching Paul Fenwick jump up on stage between the lightning talks and try to give a several minute long presentation in thirty seconds. He failed, but it was funny to watch. After the lightning talks was the closing ceremony. The main reason for this is to hand out a few awards and thank some people, but also to find out where the next linux.conf.au is going to be held. Next year, it’s in Canberra!

linux.conf.au 2012 – Day 4 (Thursday)

Thursday 19th January:

The morning started off with a keynote from Karen Sandler, Executive Director of the GNOME Foundation. This was a fascinating talk about the dangers behind having proprietary computer systems running our essential services, the things we rely on every day for our modern society. Her biggest example (and an amazingly powerful one) is the pacemaker that runs her heart (or gives it a jolt if it stops, at least). Attackers could quite easily crack into some of the modern pacemakers via the wireless signal they emit, and cause all kinds of damage (like giving a shock when it’s not needed, or failing to give one when it is). In addition, there are all other kinds of things we rely on (voting machines, car control software, etc.) that are closed source and pose a risk to us because of this. So basically it was about how we really need to move to open source systems to make the world a better place, not just because of reliability or business reasons.

Between morning tea and lunch I viewed included a tutorial on setting up a Linux HA cluster using pacemaker (with an example MySQL set-up) which I learned a lot from. I’m going to have to try out the techniques at home more, and see if I can apply them to my systems. I’m not sure how well these practises will work over wide-area networks though, as they’re primarily designed for LANs (as Michael Wheeler said to me, “It’s not redundant until it’s geographically redundant.”).

After lunch was another tutorial, this time on the basics of computer security. There are some crazy things you can do to try and break into a computer. I finally understand how buffer overflows work, and what can be done to prevent them (quite a lot, actually). Breaking into encryption was another strong topic. You can measure how much power a CPU is using to figure out whether it’s doing add or multiply, and things like that. But basically what you really need to do is maintain a much higher consideration of security while coding than you did before, and I hope this tutorial will help me achieve that.

After afternoon tea (during which I consumed a suspiciously large quantity of fruit… I must be craving healthy food) it was time for the talk given by my friend Chris and his friend Paris, titled “Android is not vi”. It’s probably the funniest talk I’ve seen the entire conference, and several delegates have mentioned it wouldn’t be out of place in a comedy routine. The talk was mostly about making the user experience on Android better, but with a surprise ending: the general principles apply to pretty much any user experience design, because they’re all the same. Not making the user think is pretty much the key.

Then of course, Thursday night means the Penguin dinner. This is the opportunity for the speeches and the presentation of the Rusty Wrench award, named after Rusty Russell who started running the conference back in 1999. This year it was given to Mary Gardiner for her services to the Ada Initiative, amongst other things. This is very well deserved too, as there is a nearly 20% female attendance rate at LCA this year. The meal at the dinner was also excellent. I had a pork salad starter, the chicken thigh main, and then an amazing mango cake dessert (pictured). After dinner, we received a speech by Senator Scott Ludlam on the surveillance state that Australia is turning into. 250000+ people have had some sort of police surveillance performed on them in Australia last year. Scary stuff, and I think it will only become scarier once we have heard what Jacob Appelbaum has to say in his keynote tomorrow morning. All in all, a great event.

linux.conf.au 2012 – Day 3 (Wednesday)

Wednesday 18th January:

Quite a few more talks today; seven to be exact. The first was Paul Fenwick’s keynote address on “All Your Brains Suck”, which is about ways of hacking and exploiting the human brain. It’s fascinating stuff, even if I had seen the talk before (when he visited UTAS and gave the talk to TUCS). I now know that scary but logically safe places (like rollercoasters) are good places for dates. And that priming people with slow words (like ‘elderly’) make them walk slower. Utterly crazy stuff, just like the speaker of this talk. 😀

After morning tea, there were two talks on filesystems. The first, on btrfs, was given by Avi Miller, who was an excellent speaker. btrfs seems like an amazing piece of kit. I had always thought that you couldn’t do much with a Linux filesystem; but I was shown to be wrong. btrfs does file-level RAID, so you can how many backup copies of a file you want on the disks. It also does copy-on-write, so you have backups into the immediate past as well. The next talk was on XFS, given by Dave Chinner, and he showed how the performance of XFS has jumped up in recent years, as good as btrfs (sometimes better) and they both far outstrip ext4 now (no surprise, ext4’s underlying technology is about two decades old).

The first talk after lunch was on Ubuntu’s ARM ports, and how development is progressing for the various ARM platforms (of which there seem to be hundreds). The most interesting thing for me was the coming-of-age of ARM servers, which consume far less power than x86 servers (a good thing for the environment) while doing a similar workload. Because most services don’t require CPU-intensive workloads, we can save even more! The second talk after lunch was on adding millions of watchpoints to a Linux system. Most Linux systems currently only support 2 or 4 watchpoints, and this isn’t enough for good security analysis. So they added a driver to Linux that steps through instructions when an out-of-memory page is accessed (via virtual memory). Clever stuff, even if most of it went over my head.

After afternoon tea, I watched two talks on bootloaders. The first, on using Linux as a bootloader (given by Peter Chubb), was interesting, though was really only useful as a method in a small number of cases. Most systems are better off sticking to something like GRUB. Speaking of GRUB, the second talk (given by Josh Triplett) was about porting Python to GRUB, and the resulting project, BITS. This seems like a fascinating thing to do, mostly because it blurs the line between what an application is and what an operating system is. This is a line that Emacs has been tiptoeing on for some time now, and it is nice to see it has a friend.

For dinner today, I attended the unprofessional delegates networking session (UnPDNS) which is held at the same time as the event for the professionals. We had a barbecue, which gave me a great opportunity to hang out with some of the other cool people who didn’t buy expensive tickets. 🙂

linux.conf.au 2012 – Day 2 (Tuesday)

Tuesday 17th January:

Day 2 was full of a lot more great talks. First up in the morning was Bruce Peren’s keynote address. He talked a lot about trying harder to maintain the ideals we’ve worked for in the past. He says the fights we fought in the past, using our moral high-ground to our advantage, we might not win now because we have business groups (like Ubuntu and Redhat) speaking on our behalf… and businesses always have to put profit first.

After morning tea I saw two great kernel-related talks, the first by Jonathan Corbet and the second by Mathew Garrett. Jonathan basically gave a rundown of the Linux kernel development work that has happened over the last year, including the release of Linux 3.0 and for the first time ever, a kernel release having less source code in it than the last one (due to some cleanup work). Mathew Garrett gave a really impassioned talk on the good and evil of EFI. From what I gathered, the runtime services stuff offered by EFI is a great idea with a half-arsed implementation. On the other hand, the secure boot offered by EFI threatens to make open-source deployment to normal users a right pain in the arse… if it is possible at all. That’s a bit of a worry.

After lunch I watched a talk by Greg Banks (who works for Opera) on renovating old source code to get it up to scratch with modern systems. His examples came from the Cyrus IMAP server, and there was a heap of great tidbits of information there. The second talk after lunch was given by Robert Mibus from Internode, about how they are implementing reverse IPv6 DNS mappings for their customer. With a possible 4TB of mappings for each customer, they have to generate them on the fly… but no existing DNS server did this. So they wrote their own. One thing I was very interested in, being an Internode customer, is that I can request to get IPv6 reverse mappings delegated to my own DNS servers; something I have already put in a request for.

The second last talk for the day was about moving large amounts of data and essential services from one datacenter to another with no loss of downtime… an impressive feat! Given by a team from Mozilla, it detailed how they prepared for and moved thir crash reporting system from San Jose to Pheonix. The last talk I saw was given by Sarah Novotny about caching databases, and how the many caches on a system can sometimes work against each other. It covered performance benchmarking and monitoring as well, just to make sure everything is running fine.

Dinner this evening was at the Irish Murphy’s we visited the night before. While a bit unadventurous, I was with a group who hadn’t been there before, and the food was still quite excellent. I was happy. On the way back to the hotel I got to do a bit of train spotting as well, which made me quite happy. 😀

linux.conf.au 2012 – Day 1 (Monday)

Monday 16th January:

Chris Neugebauer waiting for the conference to begin...

Monday is the real start of the conference. First thing to do was register and pick up the conference badge and bag. While eating breakfast (a lovely cooked breakfast of bacon and eggs for me!) I looked over the provided schwag to see what free goodies had been obtained. For most of the delegates (including me) the item of interest this year was a Freetronics Leostick, an Arduino development board in an absolutely tiny form factor. Then it was time for the conference proper to begin!

First of all was a welcome before morning tea. The sponsors and volunteers were thanked, and the conference rule #1 was revealed: Be Excellent To Each Other. So much awesome. After morning tea (where I filled up on caffeine like a camel on an all-nighter) it was talk time. Some highlights from the first day:

I particularly enjoyed Adam Harvey’s talk on migrating to PHP 5.4. His style of speech is very fun to listen to, and you always feel like you’re learning some kind of juicy gossip of how the PHP team makes decisions (which, of course, you are). I know now not to use the PHP functions that I didn’t use anyway, because it seems about half the language is being deprecated. Luckily it is the right half, including the much-hated magic quotes feature.

Rob Thomas gave two talks. The first was on building a redundant PBX system using Asterisk and a heap of BASH scripts. It seems pretty cool, though unfortunately I have no use for such a thing. Michael does though, and he almost immediately tried to get a quote off Rob to install the system at his work. The second of Rob’s talks was on common mistakes made when trying to create a highly-available Linux cluster. This tied in with his other talk, still including the same PBX system as demo material, but was a bit more geneal. Both were highly entertaining and hilariously hilarious (yet, that hilarious!) and only involved me getting rickrolled once. Oh, and those talks were where the picture of the My Little Ponies came from.
Peter Serwylo gave a talk on performing static code analysis on PHP code. He showed us a cool program written by a german bloke to do this (both names I have forgotten however :(). This really piqued my interest; I’ve heard a lot about static code analysis but really failed to see it progress any further than trying to correct variable names or tell me when I have the wrong class member. This trawled the entire function call tree and tried to find security flaws (most of which end up from being non-escaped characters passed to database queries). Very informative. That talk, along with a few of the lightning talks from the open programming miniconf, really want me to do more investigation into what programming languages and compilers can do.

Guinness Pie — The Guiness Pie I had for dinner. Yum.

After the day’s talks it was time for dinner. But what to do for dinner? An Irish pub is never a bad idea. Ballarat’s is Irish Murphy’s, and they do a very good guinness pie. It was amazing. After that, back to the dorms for some blog post writing and some much-needed sleep.