25th July 1999 Archive
Browse by publication date, or search the site.
The mysterious case of the non-existent press release...
Microsoft has been throwing its weight around in the wireless and cable TV businesses, but if a curious snippet that briefly escaped onto a wire service last week has any substance, the company clearly intends to go a lot further. According to a brief Dow Jones posting early on Friday, "In a press release Thursday [Microsoft] said it plans to develop a high-speed wireless network to play music or videos and display photographs." Fascinating stuff, particularly as Microsoft didn't issue a press release saying this on Thursday. Or even on Wednesday or Friday, come to that. But if Dow Jones reports a Microsoft story that doesn't happen, it's still interesting. The Dow-Wall Street Journal operations have a remarkable facility for obtaining advance press release information from many companies, and their relationship with the printer in Microsoft's press office often looks particularly symbiotic. So it would seem likely that there was a press release, but that for some reason it didn't get issued last week after all. The fact that Dow didn't follow the story up with expanded versions (as it usually does in these cases) lends weight to the theory that MS made a last minute handbrake turn. But why? Well, developing a "high-speed wireless network" from scratch would be fiendishly expensive, and even if Microsoft could afford it, technically the company couldn't go it alone. It already has a joint venture with Qualcomm, Wireless Knowledge, has a vision of "information access any time, anywhere, from any device," and has been buying wireless outfits it deems strategic in the wireless arena. But the component Microsoft doesn't have at the moment is the network itself, and that may be why the press release (if it really was a press release) got pulled (if it really did get pulled). Wireless Knowledge has been stitching up deals with Metricom, which is a possible partner for Microsoft here. Metricom is owned by Microsoft co-founder Paul Allen, and runs a wireless Web service in various parts of the US. It's currently testing 128k systems, so could qualify, and although its reach isn't massive, the network could be expanded, and Paul Allen's large US cable TV interests could be helpful here. Or failing this, there's AT&T, which is already a Microsoft partner in cable, and has extensive US wireless interests. Or Sprint? Or even Vodafone-Airtouch? But the trouble with the cellular outfits is that their broadband roadmaps are longer-range than Microsoft would be likely to want. Something's certainly brewing, so maybe if the Ts & Cs can be thrashed out, we'll see an announcement next week. ®
Business 25 10:50
But this approach could turn into the road to hell, if it's pushed too far
Microsoft has issued a limited developer release of Millennium, the next version of Windows 9x, but early reports of its content raise as many questions as they answer. Microsoft has said that it will not include "legacy" support in Millennium, and this certainly seems the case with the code issued this weekend. But that code is a combination of Windows 98 and Windows 2000, so as yet it's not clear which way Microsoft intends to jump. As we've said here in the past (Millennium code due), the simplest route for Microsoft is to rip out support for legacy hardware while avoiding getting involved in heavy engineering on the Win9x kernel, so although some sources think the company intends to remove Dos support from Millennium, we still tend to think it will back off from that. The current suggested schedule for Millennium is for a beta 1 in early September, and shipping in Q1, and that's surely too tight for much radical work. But this weekend's developer release does remove Dos support - sort of. Betanews.com reports that the code is a "partial merger" of Win98 and Win2k code, and suggests that as Win2k code is already non-legacy, this is a good starting point. But although this means there's no Dos mode option and no command prompt, Dos is still in there, so if the real job is to be done, it certainly hasn't been done yet. Real-mode drivers, however, won't work. You can see why these might be viewed by Microsoft as steps towards greater stability, but you can also see the inherent contradictions in what Microsoft is (apparently) trying to do here. It abandoned plans for a consumer OS based on the Win2k kernel earlier this year, and Millennium is part of the substitute strategy, the other leg being Neptune. Win2k has major problems running existing Windows games, and the earlier consumer project would have been likely to have the same problems. Hence Millennium, the Win98 kernel based OS for next year. The next version, Neptune, is intended to be based on the Win2k kernel, but Microsoft has contingency plans for a further Win9x-based release, just in case Neptune doesn't hit the spot. So although the addition of Win2k code to Millennium might aid the convergence plans and might help pave the way for Neptune, it's difficult to see how the obstacles that postponed convergence could have vanished in six months. A more likely outcome, we think, is that Microsoft is currently trying to face both ways, and will take a decision when it releases beta 1 in September. And the way the company has recruited testers for the developer release may provide some clues about what it's up to. It recently invited potential testers to fill in a questionnaire, and used this to decide whether or not they'd be allowed onto the programme. This has annoyed quite a few long-standing testers who didn't get the gig. But it means that Microsoft is being extremely careful about the demographics of this first test cycle. So maybe it wants to get a pretty accurate prediction of likely reactions to the loss of (or at least the hiding of) Dos, and get a gauge of how far it can go in removal of legacy hardware as well. It could of course still go both ways, because the next generation Wintel platform (Wintel PC2001 roadmap) roadmap gives scope for a second bite of the cherry in a year's time. The joint PC2001 spec, an early version of which was released last week, provides a design guide for PCs conforming to the Easy PC Initiative. To some extent Millennium is intended to be the OS for Easy PC, but PC2001 doesn't get finalised until February of next year, and machines based on it won't be out until towards the end of the year. So if Millennium ships in Q1, it'll actually be aiming at platforms based on PC99/PC99a. That argues for a more modest piece of work now, moving swiftly into a Millennium Plus One ("Second Edition?") around Q3 2000. Yes, we know this is only supposed to be a "contingency," but the need to roadmap in sync with PC2001 means it's almost certain to ship. ®
Business 25 12:16
Workstation powered consumer models, and Dos could have its air supply disconnected
Wintel's plans for the next generation of PC will become clearer next week, when the joint Intel-Microsoft PC2001 System Design Guidelines come up for their first public review at a two-day meeting in Redmond. As we noted last week, PC2001 is aimed solely at hardware platforms running Windows, but a close inspection of the first draft reveals a lot more the alliance's co-operation plans. Strange really, considering how much they still hate one another. PC2001 consists of general guidelines for the design of PC systems intended to ship towards the end of next year, and it also contains a chapter specifically on the Easy PC design, which is envisaged, effectively, as a new category of consumer PC. In terms of look and footprint the resulting machines will be a lot like the Concept PCs Intel showed last November. Machines based on these were scheduled to arrive this November, but if any actually do, they'll be largely cosmetic jobs - iMac knock-offs, you might say. The bottom line here is that the combination of hardware and software features won't be ready in time for them to be particularly easy to use under the covers. As an aside, if you want to know why Concept PCs may be a little delayed, you needn't look much further than Windows 2000. When Intel first mooted them at IDF last autumn, Win2k was shipping in H199, and the consumer OS based on it (which would be the one for Concept PCs) was to ship shortly afterwards. But back to PC2001. Presenting it a little earlier this year Intel evangelist Steve Barile specifically said that the PC2001 schedule had been altered to accommodate a silicon development lead time of 18 months. When we read this (PC2001 released) we suspected a closer integration of Intel hardware and MS software, and a look at the detail of PC2001 makes it look like we were right. Barile also stressed that the guidelines "should avoid surprising the industry with new requirements previously not disclosed, not commercially viable, or not supported in the OS." That last, we can presume, was a reference to Microsoft's laggard support for USB in previous operating systems. Legacy removal is the most important aspect of PC2001, and the silicon development lead time will in part be caused by the amount of board-level hardware development needed to support this. But for Easy PC specifically one of the things MS and Intel propose is "high functional integration on the motherboard." One's thoughts naturally turn to new highly integrated motherboard and System on Chip designs from Intel at this juncture. These should be pretty interesting, as the Easy PC spec requires a 500MHz CPU and 128Mb memory, which is equivalent to the workstation requirements of PC2001. Watch for these being trailed at Intel Developer Forums in September and early next year. These designs will also include "modular build to order/configure to order options." So Intel intends to implement its building block approach with multi-flavour motherboards at the OEM level, we'd guess that means. The Millennium input Microsoft software support for both this and legacy removal will be important, but the guide shows that Redmond doesn't quite have its act together here yet, and therefore plans are still in flux. It puts forward a requirement that MS-DOS shouldn't be used by system builders to run any device or software, but says that "this requirement is still under review." That is, they think they want to make this a requirement, but will back off if enough people shout at them. Easy PC machines will definitely ship without legacy hardware, but the software position is a lot fuzzier, and there's a fascinating section on what Microsoft would like to do, but might not: "Microsoft is investigating the development of non-retail [i.e. OEM] versions of Windows 98 and Windows 2000 operating systems that will support PC systems that do not use legacy components such as Super I/O, the 8042 controller, and MS-DOS." Note first that these operating systems being investigated are obviously Millennium and Neptune, and second that PC2001 was published just days before Microsoft shipped the Millennium developer release. So Microsoft signed-off these doubting words after it started briefing people about legacy removal in Millennium. Now read on: "Supporting system features being proposed for the next OEM service release of Windows 98 (Windows 'Millennium') include the following: All Windows dependency on Super I/O and 8042 eliminated. USB HID detection during boot. Native support for more USB device classes, including printer, storage and broadband modem plus serial-to-USB and parallel-to-USB mappers. Native S4 support (suspend to disk) with operating system recovery in under eight seconds, and S3 resume in under five seconds. More device support through ACPI. Mechanism to return to a previous known, good, configuration." That's probably a pretty good picture of what Millennium is supposed to be about but note that "the above list should not be considered a commitment from Microsoft. The feature set for the next release of Windows Millennium will be known before the completion of PC 2001 guidelines." So the amount of legacy removal is still in flux, and there probably is room for that Millennium "SE" update we suggested earlier today. ®
Business 25 15:19