I spent all of last week at Interop/MBX New York, which hopefully explains my absence from these pages. This event is a tough one in the sense that there's a huge amount of running around (I co-chair the wireless and mobile sessions, which are known as Mobile Business Expo, or MBX), and holding the event in New York means that (a) one spends a lot of time going from hotel to the Javits Center to other offsite meetings and (b) one ends up needing a great deal of coffee to compensate for a fundamental lack of sleep. BTW, the Javits Center still has a certain dinginess about it, but renovations are underway.

So after catching up on said sleep and thinking about all that transpired, I want to focus on two questions raised in two different conference sessions that are more than significant to the future of wireless in the enterprise, but which at present do not have clear answers.  The first of these concerns the architecture of enterprise-class wireless-LAN systems, which seems to be the last real differentiator between differing products. All enterprise-class systems, after all, can support lots of APs, and everyone will have a .11n product shortly. Perhaps management-system features, extremely important but consistently underrated as such, will eventually be the only differentiator that matters. But architecture certainly is today. How "thick" an AP? Should APs be able to directly forward traffic without going through a central controller? Should a controller be required at all (although it's a given that a management server or appliance most certainly must be)? There were excellent arguments on all sides of this issue, but we're left with having to rely on vendor claims regarding net benefits (performance and otherwise) because it is simply too difficult at present to compare systems empirically. I remain hopeful that what I call virtual benchmarking, using test equipment common to engineers but not (yet) common to enterprise planning and operations types, will be the answer, but we're not there yet. I'm planning another series of tests along these lines; stay tuned.

The second issue relates to how converged solutions will ultimately be provisioned - by a carrier, or by the enterprise directly. Many have assumed that mobile-to-mobile convergence, wherein the handoff of connections between two or more radios is implemented, must be done by a carrier using IMS or UMA, or perhaps another technology going forward - there are many efforts underway here. While a carrier-centric approach seems to be the ideal, there are two key drawbacks, apart from not having a single standard at present. The first of these is when the carriers will deploy enterprise-class solutions (T-Mobile's pioneering hotspot@home notwithstanding), and what these services might be. And, of course, the fundamental danger inherent in a carrier-provisioned solution is being locked into a single carrier, which most enterprises avoid often out of necessity. No carrier has services everywhere, after all.

The other approach is to install a gateway within the enterprise, just like another switch or router. I've tried a number of these, and you can buy such a solution today from such vendors as Agito, Aruba, Divitas, FirstHand, NewStep, Siemens, Tango, and Varaha, plus likely a few more, and Cisco in six months or so. These can work surprisingly well (although I've not tried all of them yet), with essentially no handoff latency and excellent voice quality, along with a broad array of both voice and data features. I personally believe that this enterprise-centric approach will be the most popular for a number of years into the future, as the carriers continue to grapple with spectrum-auction and network-buildout costs.

And yet the carriers themselves would likely be very well-served (yes, I'm serious here) by adopting the enterprise-centric approach to convergence as their own. What they would do is to OEM or otherwise distribute an enterprise-centric solution under their own brand, thereby locking in the customer and displacing the wireline provider - all without modification to (read: major investment in) their own infrastructures. And the tight coupling would result in even faster handoffs and better load balancing all around.

I discussed this idea with a speaker from Sprint and he said he'd not thought about this idea before. Which might be his way of being polite in the face of utter lunacy, and I can understand how this idea might be perceived as such. I'm willing to bet, though, that at least a couple of larger carriers will go this route over the next few years.