The Atlanta Linux Showcase
One of the benefits of living in Atlanta was that I didn't have far to go for the Atlanta Linux Showcase. This year's luminaries included Donald Becker of Beowulf fame, Lars Wirzenius, Dr. Cowpland of Corel, Allen Miner of Oracle, Richard Stallman, founder of the Free Software Foundation, Eric Raymond, the current media favorite, John Blair, the Samba book author, Sam Ockman of Penguin Computing, and of course, Jon ``maddog'' Hall and Phil Hughes. Linus himself couldn't make it this year, I think because of his busy schedule.
GNOME and Other Presentations. I've mentioned some of the speakers already, but their talks were terrific too. Because most of the talks took place simultaneously with other talks, it wasn't possible to hear all the speakers in their entirety, which is how I like to listen to them. But I did get to hear and meet some of the notables like Miguel de Icaza of the GNOME (Gnu Object Model Environment) project.
The GNOME project is a desktop design and object model, indeed, an environment specification, that many Linux resources are dedicated to right now. Probably the largest ``complaint'' Linux has faced over recent years is that ``it's only for scientists and uber-geeks.'' There's been a pretty good reason for this complaint; it's partly true.
Before Linux can really win the desktops of Ma and Pa Kettle, there has to be a large effort to simplify the day-to-day use of Linux. GNOME is about solving this problem. There are many window managers out there, and many handy tools, but they don't work alike, and they don't always work together smoothly. By defining specifications required for GNOME, the huge Linux programming community has again shown their enormous ability to cooperate over global distances. The Red Hat Advanced Laboratories are hard at work on this project, among many others. And other desktop projects can also be GNOME compliant, so Ma and Pa's desktops can interoperate peacefully and easily, irrespective of one's individual tastes in ``look and feel.''
Lars Wirzenius spoke about the implementation of CVS, which is a terrific product for keeping directory trees of source code organized for very large projects being worked on by developers over the Internet. Imagine many thousands of lines of code being worked on simultaneously by programmers separated by oceans and languages and all the different ways of doing things, and expecting the source tree to be intelligible to one another. If that doesn't sound like a joke, I don't know what does. But CVS helps it happen, and Lars Wirzenius should know, because he's been around on many very large projects for a long time.
Allen Miner is the VP of Strategic Business Development for Oracle. Dr. Cowpland is CEO of Corel. They spoke on how their companies are taking the Open Source plunge for Linux and joining Netscape and others. Actually, neither will be licensing their software under copyleft, but they're interested in potential business models for cooperating with the Linux and Open Source community. I didn't get the details, but as I understand it, they're both working on some way to get the benefit of outside programmers' input without giving up all the profit potential.
Open Source business models are still in their infancy. It's only been six months since Netscape released their Navigator source under the NPL, and large companies have to think carefully before taking the plunge. There are many advantages to Open Source business, but it's still experimental for large companies at this stage, and even for small companies.
Scorched Earth Software. Speaking of Open Source business models, it is also proving to be an effective defense against predatory acquisitions of small companies by big companies. The large company tactic has been to claim some sort of legal infringement on their intellectual property by some small company whose property they want to acquire. The large company's lawyers have near unlimited resources, sufficient to ``tie up'' the smaller company's revenue stream and effectively make them go out of business or be acquired. It's dirty pool to the max, but welcome to America.
The ``scorched earth'' scenario is where the small company says, ``If you insist on denying my ability to profit from my intellectual property, then I'll remove the chance for ANYONE else to profit from it directly: I'll release the source code under the GNU Public License.'' This way, small companies being threatened by hostile takeovers from larger companies still have a few extra cards up their sleeves. By copylefting their software, they prevent its acquisition by the coporate bullies. The small company can still sell value added service and related ``bundled'' tools with their GNUware. This works because the smaller company is nimbler than the bigger company and can reposition themselves faster. They are more innovative and can adapt more creatively the software they're most familiar with.
It's sad that it has to come to this, but some big companies take all the fun out of the ``free'' market!
Saint IGNUcious. I had a wonderful time at Richard Stallman's talk. What many people don't realize is that Stallman is responsible for most of the free software surrounding the Linux kernel that Linus Thorvalds developed. Stallman had been working on a project called The Hurd, a different type of Unix kernel concept, before Linus developed his kernel. In fact, Linus never thought his ``hobby'' would amount to much because The Hurd would be ready ``any time now''; and, Stallman has said that had he known Linus's kernel would have been available and so good, he never would have started working on The Hurd. But, since so much work had already been put into it, Stallman and the GNU contributors will see the Hurd project through.
The ``Saint IGNUcious'' title befits Stallman because many consider him to be a free software purist. He uses only open source and completely free software on his computers; he uses NO proprietary software at all. He walks his own talk. He urges all people to use only free software on their machines, and he gives sermons on why doing anything else is detrimental to society. Basically, preventing anyone from helping their neighbor (by sharing software) encourages them to ignore their neighbor in need. He considers signing a non-disclosure agreement (NDA) a type of sin; programmers are like software doctors, and signing NDAs are like breaking the Hippocratic Oath.
I had heard all this before having ever seen Saint Richard, but seeing him put on his angelic disk platter (which looks remarkably like a halo) and donning his robe and scepter, then reciting the GNU prayer, is a completely new experience. He is quite the entertainer. And he has an important message.
Cool Stuff. One thing I love about events like this is all the free cool stuff and unique concept gizmos. One thing was this Beowulf cluster that took up about as much space as a stack of pancakes at IHOP. At the Corel booth, there were 10 small motherboards connected at their 4 corners by insulated posts and connected to a fast switch via Gigabyte ethernet. There was no case; you could see each Pentium 2-400s (I think) on each board. Ten processors all working together as one small supercomputer for all to see, and they were running Linux.
There were other Beowulf clusters on display, usually in large server-type boxes, more like you would expect to see. (Usually, Beowulf clusters are in racks of computers against a wall or something, but many different configurations are possible.) They were smaller, production-like clusters with fewer numbers of boards than what normally make headlines or attract attention, but they were clusters.
Seeing these was at least as good as seeing all the nifty goodies and free CDs that abounded. Even though I never imagined there were so many ways to wear penguins, I also never imagined that small supercomputers built from commodity hardware would become so practical.
The Global File System. One of the technical challenges associated with supercomputer clusters is their filesystem. NFS (network file system) is an old Unix protocol that is old and frail, as well as having numerous security problems. NFS with Beowulf is an oxymoron. So several solutions have arisen for a next-generation filesystem that will meet the needs of clusters. GFS (the global file system) is one solution. I'll be writing about alternatives to NFS for clusters in future articles, because it's one big example of where the Open Source development model blows away the commercial development model in sophistication and reliability.
Another alternative I found fascinating was OPUS (Open Parallel Unified System). This is a newer project that will probably fill a niche where GFS leaves a gap. GFS runs best on expensive hardware, while OPUS is designed to run best on commodity hardware. OPUS is brand new, and GFS has been in development longer; it will be truly cool if the two projects can cooperate together to cover both high and low end cluster implementations.