On the flight home from VMworld, I reflected on the sponsor's aggressive yet schizophrenic approach to software portfolio expansion. Many additions -- such as better management software (the vCenter suite), more sophisticated virtual networking (the hypervisor vSwitch), and multitenant "virtual data center" encapsulation technology (vCloud Director) -- are smart outgrowths of VMware's cloud infrastructure vision. Others are real head-scratchers. These anomalies typically share a common theme: a focus on client devices and end users, not servers and administrators.

There was no mistaking VMware's intentions. The new initiatives got prominent billing in the two keynotes, first by CEO Paul Maritz and later by CTO Steve Herrod, and at VMware's floor exhibition ("booth" doesn't do it justice). The implications for enterprise mobile devices are still unclear, but have no doubt: VMware sees smartphone and tablet dominance as a central battleground in the war for IT infrastructure hegemony.

The most interesting of VMware's forays into client computing -- theoretically promising, but realistically perplexing -- is Project Horizon, a bifurcated effort that's part mobile application manager, part mobile hypervisor. It's designed to do for mobile devices what virtual machines and application streaming software did for the virtual desktop. Conceptually, the idea of mobile VMs has logical appeal, but there are a few (quite a few, actually) niggling details that demonstrate a significant clash between theory and reality.

First, the mobile hypervisor runs only on Android -- actually, only the tiniest subset of the platform, a single model of one LG device, although VMware announced an agreement with Samsung that will expand the menu in the coming months. Why only Android so far?

It has more to do with technical challenges than market dynamics. In a meeting with VMware's director of mobile product management, Hoofar Razavi, I asked about the implementation. It turns out that porting a Type 1, bare-metal hypervisor to various mobile devices is virtually impossible given the vast hardware and firmware differences across platforms. Thus, Horizon is a Type 2 hypervisor (meaning it's hosted). Still, most mobile kernels (Android being the one exception) place more stringent constraints on application behavior compared with their desktop cousins; for example, it's impossible to implement something as common as a malware scanner on iOS. Realistically, that means it's inconceivable that VMware will be able to port Horizon to the other half of the smartphone platform duopoly without Apple's approval and assistance -- a preposterous prospect. And without the iPhone, VMware's vision of morphing the BYOPC (bring your own PC) model into BYOMD is dead on arrival.

Furthermore, despite dramatic improvements in mobile device hardware over the past couple of years, Razavi's claim that today's Android devices have such a surfeit of CPU horsepower that running a Type 2 hypervisor won't impose a noticeable degradation in performance doesn't square with my experience. After more than a year using a Droid X, I've come to the conclusion that Android suffers from the same "bit rot" syndrome that afflicts Windows, inducing a gradual erosion in performance, responsiveness, and reliability over time. The new phone that once instantly responded to screen swipes and button presses slowly wakes to life, often taking seconds just to come out of the screen lock, and regularly hiccups handling routine gestures. Sure, today's dual-core mobile CPUs will help, at least initially, but count me skeptical regarding Android's maturity, robustness, and capability for withstanding the demands of a hypervisor and another OS instance.

More intriguing is another technology VMware previewed. Code-named Project AppBlast, it does application streaming over HTML5, greatly simplifying delivery of remote, non-native apps to mobile devices. This essentially bypasses device OS restrictions -- I'm looking at you, Apple -- and fittingly enough, VMware demoed AppBlast on an iPad. This is a very elegant way of delivering VDI in a device- and OS-agnostic way to smartphones and tablets.

Of course, sophisticated browser apps aren't new, and VMware is playing catch-up in the HTML5 application sweepstakes. Recall that Amazon recently pulled the same trick with its Kindle reader (a.k.a. CloudReader) to get around Apple's restriction on in-app purchases, and LinkedIn quickly followed with its own HTML5-based mobile client. An ex-Apple employee has even launched a company, SproutCore, whose goal is making HTML5 the new standard framework for "native" cross-platform apps. In fact, there's already an HTML5 app store, OpenAppMkt. So when it's eventually released, VMware's AppBlast technology might make it easy to stream Excel to an iPad, but will anyone care?

It's interesting to speculate about the root cause of VMware's split personality. Who decided to give the company's core focus and expertise, virtualization software and associated infrastructure management tools, co-billing with its end user and application platform software, which together account for only one-seventh of VMware's license revenue and likely even less of its services billings? Why was Maritz so eager to join Steve Jobs in proclaiming the beginning of the post-PC era?

Maybe VMware's client fixation has more to do with personal psychology than business strategy. Dreams of what might have been at Microsoft seem to torment Maritz, like a wound that won't heal. This might explain why VMware's "post-PC era" sanctimony rings hollow. But driving VMware to act like a formerly scrawny grade-school kid who grows into the high school bully looking for a fight is no way to run a company.

Unlike Apple, which truly has an end-to-end path for delivering content over a platform ecosystem spanning a growing variety of both mobile and fixed devices, VMware seems to just be looking for ways to undermine the PC, and by extension, Microsoft. To me, this confrontation isn't born from a position of opportunity and strength, but out of belligerence.

Whatever the origins, Microsoft is returning fire on VMware's turf, using VMware's recent licensing miscue as an opportunity to woo customers to Hyper-V. The "be careful what you wish for" adage rings true here. By foraying into markets and technologies outside its fundamental expertise, VMware is not only playing in areas where it doesn't, and can't, set the agenda, it risks distracting itself from its core markets. The enterprise data center is VMware's to lose.

Furthermore, while its upside is limited by the behemoths that actually do control the client, VMware's downside is waking sleeping giants that could ultimately undermine its dominance as an enterprise infrastructure and cloud services linchpin.

Along with version 5 of its vSphere virtualization platform, VMware introduced a new licensing policy that was roundly denounced by users and industry pundits alike. To determine users' plans for vSphere 5, InformationWeek asked 410 business technology pros about their intentions for virtualization in general and VMware in particular. Download the report here.