Ben Slivka is one of the finds of the Microsoft antritrust action - he turns a mean quote ("grow the polluted Java"), but he has a mind as well. Back in May 1995, Slivka produced a long, carefully thought out memo that showed he'd grasped where the Internet was headed, and proposing strategies Microsoft could use to benefit from it. Slivka's memo is entitled The Web is the Next Platform (no messing about here) and is DoJ exhibit 21.

From the point of view of killer incriminating quotes it's thin stuff, but it shows where Microsoft's thinking was at in mid-1995. Slivka is of course aware of how Microsoft can leverage its way into a dominant position, as is evident from this: "We should support all of the key Internet standards and become key suppliers of Internet technology to all comers. In parallel, we should be extending the Web with as many Microsoft technologies as possible, even if we have to modify those technologies in ways not originally intended by their designers." That's what Bill and his spin doctors would call a snippet, and in Slivka's 17 page document there's virtually no other reference to the process of modifying standards out of the public domain and into Microsoft's back pocket.

But then, Slivka is assessing what's going wrong, and trying to define the way forward, so he's not likely to be placing too much emphasis on what is essentially a traditional and mechanical Microsoft process. The most striking feature of the memo is that Slivka clearly grasps the subject far more firmly than Gates did in a memo sent the day before (Gates discovers Internet tidal wave). By 27th May 1995 Slivka is on version 5 of the document, so we can presume that if Bill Gates read an earlier version, he didn't understand it. Nor indeed does Slivka seem to understand the implications of what he's saying completely. He leads in with "The Web is an application platform (complete with APIs, data formats, and protocols) that threatens Windows."

He later talks about producing a "universal viewer" and recommends a model where "the default application model is a 'dumb' client and 'smart' server, returning to the mainframe model of the past - except that the client is much more capable than a 3270 terminal." So he identifies the Web correctly as a threat to Microsoft because of its nature as an application platform, but fails to grasp that he's basically recommending Microsoft adopt a thin client/server strategy. As we've seen since, this has potentially fatal implications for Microsoft's revenue model, and his recommendation that Microsoft "Make Windows the best client platform" seems wishful thinking when lined up beside the evidence in favour of future Sun, Oracle and IBM plans he produces.

Slivka nearly gets there by inventing the NC concept. Bill Gates himself had referred to it in his memo the day before, so it was clearly an issue for Microsoft. Says Slivka: "My nightmare scenario is that the Web grows into a rich application platform in an operating system-neutral way, and the a company like Siemens or Matsushita comes out with a $500 "WebMachine" that attaches to a TV… When faced with the choice between a $500 box (Risc CPU, 4-8 megabyte RAM, no hard disk…) and a $2k Pentium/P6 Windows machine, the two thirds of [US] homes that don't have a PC may find the $500 machine pretty attractive!" Indeed - change the numbers around a little and that's precisely the threat Microsoft faces today. But although he defines the nature of the problem clearly, he doesn't push much further.

In 1995, was it perhaps career death in Microsoft to question the ascendancy of the PC? Slivka's smart about a lot of other stuff, so it's odd that he doesn't nail this one properly. In the document, he's obviously pitching a plan to get Microsoft out of a losing strategy, and he goes in soft initially. "There was a time when we thought that we could just 'build it and they will come' with MSN, hence all the non-Internet technologies we developed (Marvel RPC, incompatible Mail & News protocols, MOSView etc.) for MSN. These technology choices were unfortunate, for (in hindsight) I think it is clear that MSN would have been much further along now if we had started from the existing Web and enhanced it."

Slivka is trying, gently, to tell the High Command that Microsoft has screwed-up big time. MSN isn't going to work, Microsoft needs to embrace the Web - but he's doing it diplomatically, so he clearly expects resistance. Here he tries to intercept some more of it: "It is possible that if Microsoft forges ahead with its current MSN plan (Blackbird, OLE everywhere, COM/DCOM etc.) and only pays the Internet lip service, we may 'pull a Windows' and end up dominating the online world. All of these other players will spend all of their time bickering about IETF standards and shipping incompatible extensions, and the Internet will end up a mish-mash of incompatible solutions."

He's good this guy, isn't he? But here's how he tries to hit the bosses' hot button: "On the other hand, it is also possible that some company will 'pull a Windows' by taking a leadership position of enhancing the Web… We have to assume that at least some of our competitors have figured out how Windows won, and are trying to recreate that strategy on the Web."

He's really good, this guy… As it happened, Microsoft's Internet strategy did turn in this period, and given the floundering Gates exhibited in his review from the same period, it does look rather like Slivka was the one who engineered the turn. It turns out this smart cookie is ex-OS/2: "When I reflect on some of our previous 'big bang' efforts - OS/2 and LanMan - the key mistake we made was not to focus on compatibility enough. With OS/2 (where I spent my first 5.5 years at Microsoft, working primarily on MS-DOS compatibility), we didn't support all MS-DOS applications, and we didn't support any MS-DOS device drivers… Regardless of all the cool, sexy features in OS/2 (multi-tasking, better graphics API, memory protection), it was not a no brainer upgrade from MS-DOS… Only with Windows 95 (where we have focused on compatibility to an amazing extent) are we finally going to enable to [sic] move customers away from MS-DOS."

As a side issue, it's worth noting that the above paragraph suggests that Microsoft took OS/2 seriously at least for a period - Slivka views its failure as a Microsoft mistake, and sees the error as being useful evidence to support his pitch. He makes the same point about LanMan: "We told customers they had to toss their existing Novell networks in order to run LanMan and they would have to accept slower performance from LanMan… With Windows NT and Windows 95 embracing NetWare, we're finally starting to gain some ground here."

Having - he hopes - got the high command's attention, he next needs to turn the tanker. The current MSN-related project, Blackbird, needs to be canned. "If we have the resource to continue pursuing the original Blackbird viewer/server design without impacting our Web publishing efforts, that's fine." He argues that by building "client, server, protocols and (some) data formats from the ground up" Microsoft will be repeating the OS/2 and LanMan goofs.

He touches on the big advantage of MSN for Microsoft, secure payment. This was what seemed to force Gates to cling to MSN, so it's a tricky one. Microsoft has to sacrifice something where it can see how it can make money, for something (the Web) that doesn't have obvious revenue streams. But proprietary is tricky. If Microsoft sticks with this, "if we force information providers to make a choice between creating great Web content or great MSN content, they may choose the former."

And he swipes at OLE. He tries to be diplomatic (internal politics) but fails. "Using OLE as a fundamental Framework adds Complexity/Size, reduces speed… If OLE were not complicated, fat and slow it would be wonderful." So MSN has to be given radical surgery, and OLE everywhere has to be shot, along with Blackbird. "A more gradual approach of starting from existing technology and enhancing it (as I'm suggesting in this memo) has proven in the past (Windows) to be more successful than a Big Bang approach (as with OS/2, Lan Manager, Exchange?)."

Slivka ends the overview section (the first half of the document) by describing his preferred Universal Client and server system, and then proceeds to explain how it will work in detail. Microsoft did turn in the direction he suggested, but didn't take the universal client to its logical, generic and not exclusively Windows, conclusion. From the point of view of internal politics that was probably impossible. ® Complete Register trial coverage