Analysis: When Linux met Wintel
What happens when a revolution with no leaders meets the Intel War Machine?
In the week or so since Intel invested in Red Hat the 'mainstream' computer industry has been falling over itself to take Linux seriously. Its prospects of displacing Windows have been assessed, Caldera has suggested that it too might be getting some Intel money, and Red Hat CEO Bob Young is predicting that six out of the top ten server manufacturers will be offering Linux by next year. As we understand it a considerable proportion of the Linux 'community' (whatever that is) couldn't care less whether the 'proper' PC industry takes Linux seriously or not, but as we hope to explain here, it's going to be impossible to ignore the process of being taken seriously, and the phenomenon of everybody loving Linux isn't necessarily going to work out for the best. The basic problem is that the PC industry business model and the Linux one are probably mutually incompatible - as a matter of fact, it's kind of difficult to pin down any kind of business model that the Linux one could be thought of as being compatible with. Giving the product away and selling the service and add-ons is understandable to business (e.g. Gillette and razor blades), but the 'revolution without leaders' aspect is alien - who do you deal with? Intel must to some extent have understood this when it put money into Red Hat, and when it joined Linux International. Membership of Linux International gives it a seat at a top table of some form (but not, as we understand it, the top table), while buddying-up to Red Hat gives the company something to negotiate with. It also gives Red Hat access to pre-release information on Intel processors, just like Intel's other strategic partners. But this is obviously a two-edged sword. First among equals? If Red Hat benefits from getting Intel confidential data in advance of the rest of the Linux world the company's probably dead meat. It will be seen by some as being sucked into being Intel's tame, proprietary Linux, by others as bidding for leadership of the Linux world - basically, if Intel set up a Red Hat relationship that was absolutely equivalent to its other strategic relationships, it would probably break the whole thing. Bob Young seems to understand this, and if Intel is really going to put money into Caldera as well it might well be that Intel at least half understands it. Intel needs to figure out a way to co-operate with Linux and share information without favouring any one outfit. But if we look at what Linux developers want, how Intel's business model works, and then try to put the two together we can clearly see big problems ahead. The Linux people want the hardware information that will allow them to produce software, and the obvious way to deliver this within the Linux model would be for Intel just to post the whole lot on the Web as soon as it's got it. But this is where the mutually incompatible business models come in. Intel is constantly changing its hardware, and uses data on forthcoming hardware as a competitive weapon. Digital sued Intel because of this, and Intergraph is currently suing Intel because of this, claiming the company cut off the information supply as part of a bid it blackmail intellectual property out of it. And a few years back Intel pushed the Pentium into the market faster than the old-guard of the PC industry expected or wanted, by the simple expedient of building the chipsets and the motherboards itself, and leveraging alliances with the companies that would go along with the approach, Dell, Gateway and Packard-Bell. These three benefited hugely, and it cost Compaq, which missed the early Pentium boom, a bundle. Intel uses its proprietary information to play hard-ball, and regards it as the key weapon in its armoury. It's willing to share that information, but there's always going to be a price - Intel might be able to evolve the model so it could negotiate the price with a couple of Linux companies, but it can't go the whole way without destroying its business model (mind you, the FTC antitrust action against the company next year may end up doing this by force). Naturally Intel isn't particularly keen on such overly brutal interpretations of its business model, but even if we take a more positive view of it, say from the perspective of Dell, we can see problems arising as the PC business tries to adopt Linux. Linux and the Wintel model In saying six top vendors will soon offer Linux, Bob Young is just stating the obvious - Intel backing together with the demand they'll have been experiencing makes it inevitable that the major PC companies will start experimentally offering Linux in the very near future. But how? Dell's relationship with Intel is just about the strongest of any PC company, and the way it puts its machines together is approximately as follows. It relies heavily on Intel's R&D, and does R&D of its own which it then leverages with Intel's via cross-licensing deals. Keeping close to Intel developers allows it to stay within industry standards (it's a given that Intel is the company that defines the standards) while influencing them in the directions it wants. The approach has been massively successful, and both Dell and Intel love it. The co-operation obviously has to cover both hardware and software, so Microsoft intersects with the process, and the result is that Dell gets totally integrated and shrink-wrapped products to market early, frequently first. We've used Dell as an example, but the approach is similar throughout the PC business. Acceptance of Wintel hegemony lets them all keep in step on standards, and lets them get shrink-wrapped, ready to go products onto the market. Now, if the PC companies are going to offer Linux, it's obvious that the business models will become more and more obviously incompatible as their Linux offerings become more and more broad range. Each of the companies will tend to want to adopt a single Linux as its standard, and it will want to have its own software and hardware developers working closely with Intel's in order to optimise the end products, and give it the ability to sell fast, reliable read-to-go Linux servers. A Linux distributor might get involved in this process, or it might not. The upside of this will be that at least Microsoft won't be allowed to play, but there are obvious dangers - a few Linux operations may get most of the pie, the PC vendors themselves could end up producing their own 'commercial' implementations of Linux, or Intel could wind up running a de facto proprietary version of Linux as the new industry standard. We're not saying this will definitely happen, so please, no hate mail. What we think we're trying to say is that the collision of Linux and the conventional PC industry is going to have odd, unexpected and unpredictable consequences. In the worst possible case scenario, we get a repeat of the Unix wars, and somehow Bill and his Merry Men emerge triumphant again. A better outcome that involves both the PC business and Linux changing is possible, but it will be tricky to manage. And yes, we do think Linux will change, whether or not the 'community' resists it. ® Click for more stories
Sponsored: Hyper-scale data management