MS roadmap for next Win9x and consumer NT leaks out
And hopelessly over-optimistic it looks to us, too...
Microsoft's post Win2k operating system roadmap is back on track - but considering the size of some of the potholes, it's unlikely to stay there. According to documents leaked to ZD mags Smart Reseller and PC Week, there's one more rev of Win9x due next year, and then the consumer version of NT for 2001. (ZD story) Regular readers will recall that at WinHEC 98 his Billness described Windows 98 as the last rev of the Win9x line, and announced that Microsoft would converge on the NT kernel, with a consumer version of NT following the release of Windows 2000. The sheer nightmare of actually implementing a viable consumer NT resulted in the demise of that plan earlier this year, and a swift upgrade of Windows 98 Service Pack 1 into Windows 98 Second Edition. By pulling the plugs on the consumer OS, Microsoft caused a deal of confusion around WinHEC 99. A swift deal was done with Intel to produce the Easy PC "joint" initiative, which gave Microsoft something to announce. But if you picked away at this one it became horribly apparent that Easy PC was an initiative Intel had prepared earlier, that Intel had been expecting Microsoft to come up with data on Consumer NT and not, er, Win98 SE, and that the addendum to the joint PC99 spec would be out late, owing to a wrong operating system type situation. Red meat on Easy PC itself has been thin on the ground since it was announced, so it's highly significant that Microsoft now appears to be telling ZD that next year's rev of Win9x (codenamed Millennium) will be "a key component of Easy PC." That might be what Microsoft reckons, but practically all of the published work on Easy PC so far has been carried out by Intel under the auspices of Easy PC's Intel-only precursor, the Ease of Use Initiative. One particularly intriguing by-way under Intel's version of the Easy PC banner leads to the Intel Hardware Implementation Guide for consumer PCs in 1999 and 2000. This guide mentions Windows 98 a couple of times, but stresses that "much of the content applies to other consumer and business platforms based on the Intel architecture." So not only is Easy PC a la Intel not dependent on Millennium, it's not dependent on Win9x either. And, says Intel, much of the technology needed exists now, and the "Easier to Use PC platform" (relation) is "to be introduced in 1999." Funnily enough, the Microsoft Easy PC announcement said hardware companies would have prototypes out by the end of this year, so if MS wants these to run Millennium it had better get its finger out. But having its 'partner' singing from a substantially different songsheet is the least of Microsoft's problems with its newly leaked roadmap. It seems clear that the new 'last' rev of Win9x, Millennium, will either be a nightmare to develop or will turn out to be just another service pack on steroids. Similarly 2001's effort, the revived consumer NT (codenamed Neptune) will be a development pig - Microsoft's confidence in it actually happening can be gauged by the existence of a contingency plan to ship yet another 9x rev in 2001 after all, in the event of Neptune failing to make the grade and/or ship on time. Microsoft is telling potential beta testers for Millennium that they should expect it to be "legacy free," i.e. that it won't include Dos support any more. That means a kernel rewrite, and doing this while maintaining compatibility with Win9x software (games being particularly important for a consumer OS) will likely cause plenty of headaches. At the moment therefore we'd say Microsoft is going through its standard over-optimistic phase in developing Millennium - later, as the deadline looms, it'll quite possibly get scaled back to that service pack on steroids we mentioned. ZD predicts a public beta for late summer this year, but it's difficult to see how that would fit with a radical rewrite. Neptune, the consumer NT, is billed to include various goodies such as a new, Web-like user interface, more meaningful error messages, self-healing and self-updating features and so on. This stuff is relatively familiar, consisting as it does largely of standard bolt-ons that might or might not make it into Millennium or various other upgrades and service packs Microsoft might ship in the interim. But the big problem for Neptune isn't going to be the bolt-ons and go-fasters - it'll be building that consumer OS on top of the NT kernel while maintaining compatibility with Win9x software. Only a quarter of Win9x games run on Windows 2000, and it's difficult to see how Microsoft can improve that figure with Millennium. Earlier this year there was talk about Microsoft shipping a kernel upgrade for Win2k during next year, and we'd expect this to be justified in part by the kernel requirements of Neptune, but the whole thing is likely to be so messy that the next Great Leap Forward OS will slip, and the contingency 9x-based upgrade to Millennium will ship instead in 2001. So there you have it. Microsoft historically has had two operating systems, Win9x for consumer (plus) and NT for business. The plan was to converge these at Win2k, but that was postponed, so we've currently got Win98 SE and Win2k (RSN). The plan now is for convergence to take place on the NT/Win2k kernel by 2001. But the likely result will be a Win2k-based business OS and a Win9x rev for consumer in 2001. Plus of course, an ongoing convergence plan. Plus ca change... ®
Sponsored: Transform Your IT Infrastructure