Opinion: How Linux could screw MS in Q4 2000
What if something else was ready to roll when MS missed the deadlines, this time around?
After a few months hiatus Microsoft and Intel seem to have got their act back together, and have produced the first draft of the roadmap for the class of PC that's intended to go on sale late in the year 2000. They've done joint roadmaps before, and it has worked to the extent that Intel has remained the dominant chip supplier and Microsoft the dominant OS supplier - but what if, this time around, it fails? Actually, the latest blueprint has a high probability of failure on the software side, and this is a clear opportunity - if Microsoft's rivals are smart enough to take advantage of it. Take a look at the kinds of machines that are already coming onto the market, and the way these have been roadmapped by Intel first, and then more recently by a reformed Wintel alliance in the shape of the Easy PC Initiative. PCs are currently getting cheaper, to the point at which they're "free" (i.e. extremely low-cost pieces of hardware that come bundled with a two to three year ISP contract). Customers are becoming more receptive to the idea of some kind of Internet access box rather than a PC per se, and an Internet access box just has to be able to browse the Web and collect the email - it doesn't have to be Windows, and as Windows is expensive and overly-complicated, quite a lot of these boxes won't be Windows. They'll be Linux, BSD or BeOS, and the customers won't really care. Next year's model That's the obvious opportunity that already exists, but if you look at how this class of hardware will develop, you can see how the opportunity will widen, and how Microsoft's opportunities for failure will increase. Last autumn Intel started talking about the precursor to Easy PC, and this led to the demonstration of a series of 'Concept PCs' in Las Vegas that November. Shortly after that Microsoft and Intel were clearly co-operating in the development of a basic platform that would allow those machines to be built, in the second half of this year. That turned out to be strike one for Microsoft. In late 98 the company intended to build the consumer implementation of Windows 2000 for what was then Intel's Easy to Use PC Initiative. As this consumer OS got cancelled/postponed, Intel's Kahneeta 'Always On' PC reference platform also got cancelled, and the two revamped their plans in the shape of the Easy PC Initiative, which was titularly joint but largely Intel-derived. That's where the opportunity for Microsoft to make strikes two and three comes in. Easy PC platforms are scheduled to go into manufacturing towards the end of next year, and they're going to need operating systems. The hardware specification for Easy PC is already pretty clear, but the information leaking out about the first impression of the software Microsoft is aiming at it, Millennium, speaks volumes about how unready Microsoft is to roll with the next generation of simplified PC. The vagueness of the references to Microsoft software in Easy PC is also significant. The basic point about Easy PC is that all (unless the designers chicken out) of the legacy hardware is going to be ripped out. From a hardware point of view that makes the PCs easier and cheaper to build, and it should also make them more reliable and cheaper to support. But although it's possible to take out ISA, parallel, serial and floppy and substitute USB and Firewire, and then still run the machine on a hacked-about version of Windows 98, it's hardly ideal. The kludge An operating system that's still ready and waiting for pieces of hardware that aren't ever going to be there is going to be slower and less reliable - the dodgy hardware might be gone, but the dodgy routines the OS goes through in association with that hardware will be a problem for as long as they remain. And then there are those bits of Win16 code still rattling around causing problems. Ideally, you'd start with a new, non-legacy kernel and build a consumer OS specifically for the Easy PC platform; this is of course was Microsoft planned and then cancelled earlier this year, and what Microsoft is not planning now. Millennium is being built on Windows 98, and will not include a radical kernel rewrite. It may well end up blocking out legacy hardware and shielding the user from Dos, but it is surely inevitable that the legacy code will continue to rattle around in there, undermining the overall integrity of the platform. Microsoft might manage to do a better job with Neptune, the next effort at consumer Win2k, but that's not likely to show up before 2001. So there's the opportunity. Hardware designs that could really use a radical rewrite of Windows will be around in the second half of next year, but Windows will at best be able to cater for them in 2001, and don't count on it. The hardware spec will be developed over the next six months or so, with sign-off and publication due in February. By that time hardware companies, cable TV outfits, ISPs and goodness knows what other kinds of companies will be even keener on slimmed-down, simplified and easier to use designs. And they'll be looking for honed, battle-ready, low-cost operating systems that are appropriate for the specification. Open source outfits that started designing now could give them what they want. Microsoft's inability to keep pace with the demands of the hardware is hardly news. Through the various iterations of the joint PC9x system design guides the company has repeatedly failed to deliver the software on time - more usually, it slips a year, even two years. In the past that hasn't mattered for Microsoft, because it's been the only game in town, so the hardware has just had to wait for Microsoft to catch up. But the roadmaps for next year's platforms are there, the hardware data is coming, and Microsoft doesn't look so much like the only game in town any more. If the rivals start now, Christmas 2000 could get pretty interesting. ® See also: Linux loses in NT tests - Mindcraft numbers still wrong? Benchmark battles - now Linux beats NT 'Hit team' drives MS plan to bludgeon Linux with benchmarks Tests cited by MS prove flaws in Linux study - Linux Today Can Linux avoid Microsoft's NT trap? 'Show you can beat NT' - MS declares war on Linux Linux needs MS Office to make breakthrough - report MS Office for Linux - dream or nightmare? Maritz on... Linux We taught the Linux boys everything they know -- SCO exec Leaked Microsoft memo outlines anti-Linux strategy Linux - now Intel stabs Microsoft in the front Analysis: When Linux met Wintel