Day traders don't rate SGI Linux distro

Linux moves noticed, but not much appreciated -- yet

If SGI was hoping to reflect some of this week's Linux fever onto its own announcement last week, the exercise can only be judged a partial success. Already this month, more SGI stock has been traded than in the preceding eleven months of this year but the price, although edging up, is short of even half of its February high.

Over here at The Register, we can't find much to disagree with the conventional wisdom that the company is imploding in a fairly spectacular fashion. But whether this proves to be terminal is another thing - Apple found itself in even worse circumstances three years ago, and is now profitable, and investors who clung onto their shares have seen their value quadruple. (Or double if you bought in the glory years.)

At least SGI it knows where it's going - and its plan B - getting Linux out in volume - makes far more sense than its NT wheeze ever did. SGI has now put all its eggs in the Linux basket: releasing key intellectual property as open source, and creating its own distribution - a value- add 'Plus Pack' on top of Red Hat Linux, a la Mandrake.

And the latter part of this - the grandly titled SGI Linux Environment --inched forward on Thursday with the release of some kernel fixes. This is something of a curate's egg, but the best of it is really rather interesting. The release wraps up beta patches from the forthcoming 2.4 kernel from Stephen Tweedie and Alan Cox, with some home-brewed patches and drivers.

Some of this, such as the spinlock metering - which is instrumentation hooks only really useful to other Linux kernel developers - was rejected by Linus for inclusion in the main Linux kernel tree. But overall SGI has a pretty good hit rate for getting its submissions included accepted so far.

Now these bits and pieces may sound like small beer compared to the really meaty stuff SGI has promised to make available under open source - NUMA support and the XFS journaled file system -- but it's not to be sneezed at.

The 64-bit file system support, support for big memory, RAID, large volumes and asynchronous I/O are all tick-list features Redmond has shouted about in its recent NT vs Linux spins. And we note that at least one of these - 64-bit file system support is a nasty kludge until the real 64-bit Windows appears sometime in 2002. The downside for SGI is that throwing these pearls into the Linux kernel stream is no guarantee that they won't simply get washed away.

And maintaining features which really should be in the kernel, but aren't, becomes increasingly difficult as the juggernaut rolls forward - that's one of the big reasons the mainstream kernel hasn't forked: it's too much trouble. So if SGI wants to take centre stage as a Linux mover and shaker - it had better start seeing that these punts keep hitting the target. For kernel patches, that means inclusion in the stable tree. For extended features, that means widespread adoption in the other distributions. With the release of XFS we'll get a better idea if SGI is going to be more than a peripheral player. ®

Sponsored: Designing and building an open ITOA architecture