Microsoft faces its Scopes moment
48 hours that will shape IT
Europe vs MS Microsoft begins the most important 48 hours in its history on Wednesday in Luxembourg. It's not exaggerating to say that the future shape of enterprise IT hinges on the decision the 13 Judges will make, based on this week's public hearings.
The crucial decision rests on interoperability, and it could lean one of two ways.
A decision to hold Microsoft to its word on server and network interoperability will dramatically enhance the health of the Windows eco-system, with a market of third party add-ons and accelerators based on Windows protocols. Microsoft is at its most productive when it faces vigorous competition, and it's least productive when it has no competition. In addition the company fears the "commodification" of its protocols, superseded by cheap and open internet standards. Creating a flourishing market based on Microsoft protocols, not internet protocols, blesses and enshrines the Microsoft way. By any rational yardstick, all this is good for Microsoft.
But a decision by the judges to throw out the EC decision will affect regulatory interventions for years to come. Microsoft will have proved, by reductio ad absurdum, that it is impossible to effectively police a software monopoly. Microsoft will have won a permanent, government-blessed shield from the rigors of the market. Legislation designed to check an unfair monopoly will have been successfully used to protect the monopolist's business.
This is even better for Microsoft, and you can see why it's fighting so hard to achieve this outcome. It prefers to pay lawyers, and public relations bloggers, rather than give engineers the scope to innovate, because the outcome is so potentially lucrative.
In this light, we need to rethink some assumptions about the case. For years, critics have suggested, Microsoft has regarded its antitrust legal costs as a necessary business expense. Perhaps the critics have had it wrong, and we should jot down the antitrust costs in the capital investment column. For if Microsoft succeeds in nullifying antitrust legislation it'll be able to congratulate itself on one of the best RoIs ever made by a technology company.
The law isn't static, or morally neutral. While this week's arguments are very much about the ethics of regulation, there's an outcome that's the antithesis of whether companies should be "free to innovate." It's really about whether any single company deserves to be protected from competition. Far from taking a moral stand against government meddling, what Microsoft is asking for is special-case pleading: it wants and needs the government's legal umbrella to protect it from competition. If Microsoft prevails in dispatching GPL-based competition in Europe, it will have won the right to a form of corporate welfare in perpetuity.
Let's hear the arguments.
Today, organizations that have invested heavily in Microsoft technology will look in vain for such products as an Active Directory accelerator from Cisco Systems. It doesn't exist. There is no thing as an Hewlett-Packard Exchange Mail accelerator. Network Appliance can't offer a Primary Domain Controller appliance. If you want to speed up mail, directory services or network administration, you need to buy more servers, each with a Windows server license, or buy a faster Windows server.
That's because Microsoft runs a motley collection of unruly and deeply obscure protocols, and it guards them jealously. The issue of building an interoperable product - such as the ones described above - hinges on how well the product works with Windows servers, and these protocols, which have evolved over 20 years in haphazard fashion, dictate how.
So prospective manufacturers face two obstacles, one technical and the other economic - and this week's arguments hinge on whether Microsoft's proposed remedy, its Microsoft Workgroup Server Protocol Program (WSPP) is acceptable.
The Commission's 2004 conclusions, which Microsoft is appealing against in their entirety, oblige Microsoft as part of its remedy to "within 120 days ... disclose complete and accurate interface documentation which would allow non-Microsoft work group servers to achieve full interoperability with Windows PCs and servers … [our emphasis] so rivals can "develop products that can compete on a level playing field in the work group server operating system market." In this respect the remedy went far beyond the Microsoft Communications Protocol Program (MCPP), created after the US antitrust action, which specified only server to client interoperability.
The technical issue is really twofold: what can an interoperable product actually do, and how well can it do it? Right now, the reason Cisco can't build an AD Tune-Up box is because it can't create an interoperable AD implementation. It can only go so far. The Samba team, which creates a GPL implementation of Microsoft protocols used by many commercial IT vendors, still lags important features. In a submission to the EC earlier this year Samba's Andrew Tridgell compared the process to reassembling the pieces of a broken jar.
"While dedicating a lot of highly skilled work just to reconstruct the smashed jar, Samba has done a great job achieving an equivalent implementation of Windows NT 4 - alas six or seven years later," he wrote.
Proprietary vendors who want to go down the official MCPP route soon find themselves handicapped. Because Microsoft doesn't want a rival creating what's called Primary Domain Controller functionality, no manufacturer is permitted to look like, walk like, or talk like a PDC. And it's impossible to interoperate effectively with a Windows PDC without this functionality. Network Appliance lost much of its potential future growth when it agreed to go down the MCPP route. It can't do server to server communications effectively.
The economic issue is whether it's worth a manufacturer building such a product at all.
Microsoft has set the baseline for revenue it wants on Windows networks. Microsoft wants every client that connects to a Windows server to pay a Client Access License, or CAL. It's a per-seat license of around $50.So Microsoft's proposed royalty schedule is a strong disincentive for manufacturers. It isn't so much the $50,000 upfront in advance royalties - which a Cisco or an HP can easily afford. It's that when users are taken into account, the costs become unattractive.
That file and print accelerator will cost the manufacturer a minimum of royalty $80 minimum per box, and a cap of $1,900 per unit. So is he offering $80 more value than a generic, off the shelf Dell? Is the IT buyer getting $1,900 more value from that box? At these prices, no one's going to build such kit.
Then there's the black hole. Cisco had a product called CNS/AD which put an Active Directory implementation on Unix. Here's the 1998 press release. It was never released - we don't know why.
The EC also heard how AT&T had developed an "Advanced Server for Unix (AS/U) which was able to run as a PDC in a Windows NT domain. Sun licensed it as Project Cascade, and sold as Sun's PC NetLink it allowed a Solaris server to act as a PDC transparently, supplanting a Windows server. AT&T had 11 licensees for AS/U but fell out with Microsoft over the rights to access to the future versions of the Windows source code. One Microsoft executive wrote, "we are going to need to get to our checkbook in a big way… We can buy this out if we need to but the [struck] might be very big. Very".
Microsoft succeeded in striking a private financial arrangement with AT&T, and the licensees lost updates to the software.
This year's fireworks have focused on how well Microsoft's documentation does its job in allowing interoperable products to be created. Microsoft has provided a barrage of specially-commissioned reports to argue that the documentation is adequate. But the Microsoft-shortlisted monitoring trustee Neil Barrett, doesn't appear to be impressed.
Alongside our special reporting from the court in Luxembourg, we'll go into much more detail on these crucial enterprise IT issues tomorrow.
Look for Microsoft to produce witnesses who argue that server to server products can be created just fine. Look for critics to pick holes in the protocols documentation Microsoft has offered.
And look out for some examples of Microsoft at its most paranoid: where "interoperable" means "they're coming to take our children!"®