January 17, 2007

Linux down under: Day two at LCA 2007

Author: Joe 'Zonker' Brockmeier

SYDNEY -- The seventh Linux.conf.au (LCA) continued Tuesday at the Kensington campus of the University of New South Wales in Sydney, Australia, with miniconferences and a keynote by Christopher Blizzard.

Blizzard spoke about interface design for the One Laptop Per Child (OLPC) project, and how to be relevant to users. He ventured into slightly dangerous territory, saying that part of being "relevant" means being able to "move the needle" -- that is, increase adoption -- and asked, "How many years now has it been 'the year of the Linux desktop'?"

Blizzard's keynote was well-received, but it contained nothing new. The most-asked question during the Q&A session that followed was "When can I buy one?" The answer is still "you can't," at least for now. Blizzard and the OLPC project are focused on getting the units into the hands of the intended audience rather than selling them to hobbyists and geeks.

Miniconfs continue

Tuesday's schedule included a full slate of miniconfs. The Debian, GNOME, and education miniconfs carried over from Monday, and Tuesday also featured miniconfs on gaming, kernel development, PostgreSQL, OpenOffice.org, and Linuxchix.

I enjoy games as much as the next fellow, so I started the day off with the Python and the BigWorld talk by Paul Murphy. Murphy's talk was primarily about why developers might prefer Python to other languages for scripting, and about how BigWorld is using Python. Toward the end of the presentation, Murphy demonstrated how he could modify in-game elements of a Massively Multiplayer Online Game in real time using scripting. It was impressive how little code was required to do something.

Rule 0: Don't panic

After Murphy's talk, I cut over to the Linuxchix room to catch "Bug Fixing for Non Programmers" by Akkana Peck. Peck's presentation was a high-level guide to ways that non-programmers can trace down and possibly even fix program bugs. "Rule 0," according to Peck, is not to panic. Instead, users should work through a problem logically; it's likely they can find a solution with patience and persistence.

Peck discussed using utilities such as grep to search through code and hunt for clues how to fix programs, and explained briefly how to walk through compiling code from source for most open source programs. Peck also talked about how to submit a patch, and how best to work with projects and maintainers.

It seemed like the audience was already familiar with the tools covered, but the audience seemed interested in how to contribute and fix their own bugs. After the main talk there was a lively question and answer session, focusing in particular on how to get patches accepted in projects where the maintainer may not be immediately receptive to new code.

Wesnoth for kernel hackers

Rusty Russell

Peck's presentation was good, but the next talk I went to was the best I've been to all week. As part of the gaming miniconf, kernel developer Rusty Russell talked about how he became involved with the Battle for Wesnoth project, and some of the challenges he'd run into when trying to add features or fix bugs in the turn-based strategy game. The room was jammed to capacity, with every seat filled and eager attendees crammed into every nook in the room that would hold a body semi-comfortably.

After several failed attempts at playing the Wesnoth trailer, Russell got it running on the projector with a little audience participation help. After that, Russell discussed some of the technical challenges he'd run into, but from a high-level perspective that was accessible even to the non-programmers in the audience.

In addition to being easy to understand, Russell's presentation had the added benefit of being hilarious. Most of the talks I've been to have had some measure of humor, but Russell is a natural presenter and had the packed room laughing it up through most of the talk.

Russell talked about how he went about trying to optimize the game's AI so that it would be faster at simulating attacks, and also how working on a open source game is different from working with the kernel: "In the kernel, we don't deal with musicians." Russell noted that musicians and artists, part of any gaming project, may not be familiar with the terms of open source licenses -- or ready to deal with allowing others to tinker with their creations.

Tea time

After Russell's talk, it was time for afternoon tea. Twice a day, attendees are treated to light snacks -- pastries, fruit, and other assorted munchables, and a selection of beverages -- for morning and afternoon tea.

As attendees descend on the snack tables, they can mingle and participate in LCA's rich hallway track. I've been spending a lot of time polling attendees to see what kind of folks are attracted to LCA and how they're finding the conference. Almost all of the attendees that I've talked to work with Linux in some capacity, and there's a pretty good mixture of sysadmin types and programmers in the crowd. The take on the conference so far is uniformly positive, with a few minor grumbles about lack of network connectivity early on. However, those problems appear to be sorted out, and most, if not all, of the colleges now seem to have wireless access, and the connectivity around the proceedings is handling the onslaught of users well.

How many Gmail users does it take to make up a Debian mailing list?

After tea, I returned to the Debian miniconf to hear a 20-minute talk by Pascal Hakim, one of the Debian listmasters, who discussed dealing with Debian's mailing lists from a technical and social perspective.

Hakim says that the Debian lists have 176,769 subscribed addresses, and 90,409 unique addresses. (Users may be subscribed to more than one list.) Of the unique addresses, Hakim says that about 25,000 are Gmail addresses. How's that for market penetration?

What's the average traffic look like? Hakim says that 1,133,933 messages were sent on Monday, which was not a heavy traffic day.

He also talked about spam problems with the list. First and foremost, Hakim says that they want to make sure mail gets through -- and if they have to choose between the possibility of hindering legitimate messages and letting some spam get through, the legitimate mail takes priority over stopping spam.

I'd hoped to see the talk scheduled after Hakim's, which was listed as "Release Monkey" by Jonathan Oxer, but for some reason that presentation didn't follow Hakim's, and we were treated to a discussion of Maemo instead by a presenter whose name I did not catch when it was thrown out. The presentation was somewhat interesting, but a bit disorganized.

After the Maemo presentation, Robert Collins of Canonical got up to talk about a problem he sees with Debian unstable -- namely, that it's living up to the title a little too often. Collins proposed that Debian take steps to avoid having broken packages in the unstable archive as much as possible. For example, if a crucial package (such as GCC) is going through a major update, he proposed that it be done in a branch of the archive until it is stable enough that packages that depend on GCC will not be broken.

This got a lively debate going, but no major consensus was reached.

The miniconfs wrapped up at 5:30. A Linux Australia meeting was scheduled right after that, and a conference party sponsored by Google was scheduled after that, which (by all accounts) was fun for all.

LCA 2007 runs through Friday, January 19, at the University of New South Wales. Thursday is Open Day, so if you're in the Sydney area but not registered for LCA, you might want to sign up and head down to see the technologies on display then.


  • News
Click Here!