TurboLinux puts its case on clustering
Suck it and see, says TurboDoc...
The keeper of the TurboLinux kernel tree, Christian 'DocWhat' Holtje has written to us clearing up the company's position to its load-balancing patch to the Linux kernel. So let the man speak: "TurboLinux is not going to fork the kernel. One little module does not fork the kernel. Lots of projects do this, including advanced RAID, USB (originally) and PCMCIA. They are not "forks" but independently maintained enhancements. "The difference is we keep our 'patches' up to date with the core kernel. We don't try to maintain the whole thing separately. If we forked the kernel, we would have to maintain it entirely ourselves. This would lose all the advantages of Open Source (large pool of bug fixers, enhancements in areas we don't have a pressing need for, etc.). There is no point to it. Frankly, I feel faint just thinking about the work that would be required." With which we wholeheartedly agree. As we suggested here, the fork "scare" was a piece of yellow journalism, aided and abetted by an analyst. We hope he was misquoted, for if not, as a Unix expert he needs a rapid refresher course in recent Linux history. What's intriguing is that TurboLinux's business model is also getting clearer. Back to the Doc: "TurboCluster's Module works just fine without our proprietary tools. It has a simple interface via /proc/net/cluster/config. If you want a GUI, documentation, support, the newest patches, then you want to buy our product. But if you only want the module, I have placed it, by itself, on a Web page. Granted, there is no documentation in the module, but the source itself, as it is released. It is a 'DIY' in that respect. But you can always download our product via our FTP site, which includes this module in the kernel RPM, meaning there are easier ways than 'DIY.'" Clear? Almost. Now before you all jump on TurboLinux, the company is only road-testing the kind of cost-based upgrade treadmill that the other distributions have been implementing for some time too. With Red Hat, for example, you can download the OS and subsequent patches for free, but pay for the shrinkwrap versions. That fee not only includes support but reflects the added costs to the producer of the packaging, and the benefits to the user of the convenience of a one-time install, and not wrapping up that leased line for an hour. But we'll wait and see how many TurboLinux-only dependent modules the Doc mentions, like the GUI, are also freely available. And then we'll see how much the community objects to, or responds to, filling the gaps in itself. But we'll also repeat the point we made last week: Linux load-balancing will eventually be part of the bigger picture, incorporating Linux/HA work, RSN. And that will come with no strings attached, we expect. So the argument, sooner or later, will be pretty academic, and our advice remains: wait for Tweedie's APIs. Anyone from Turbo want to convince us otherwise why we shouldn't? ®
Sponsored: Hyper-scale data management