Good scientists learn to question assumptions. Every so often, something you accept as axiomatic turns out not to be true. So hold on to your hats, because I'm about to question three of the most cherished assumptions in networking.First, that voice is the single most mission-critical application. Don't count on it. An IT executive recently told me that in his organization, e-mail is more critical. In a recent benchmark of IT executives, e-mail was rated as by far the most mission-critical application, with 50% rating it critical, far ahead of VoIP and call centers (rated critical by only 8.3%).IT executives are telling us explicitly that their companies are increasingly running on e-mail. Says the network manager at a major law firm: "The attorneys get what it means for e-mail to be down. Most of our senior partners charge $1,000 an hour, and we have 500 of them. So we lose $5 million if we [have] an hour of [e-mail] downtime."Obviously, both voice and e-mail are critical (you can't imagine those lawyers happily tolerating phone outages any more than they tolerate e-mail outages). But the news here is that IT executives are increasingly finding e-mail at least as important as voice - which upends the "voice is most important" argument.The second questionable assumption is that voice systems are inherently engineered to greater reliability than data networks. Wrong again.While the notion that the Internet was designed to withstand nuclear war is something of an urban legend, the distributed, probabilistic nature of the 'Net is inherently better at routing around failed nodes than the top-down, deterministic nature of the telephony network. Want proof? I was in Manhattan on Sept. 11, 2001. Though both the landline and cellular voice networks failed, I was able to send and receive e-mail and instant messages throughout the morning.Which brings me to the third assumption: that the lack of availability of E-911 capabilities is a serious and unsolved problem that's rightfully holding back deployment of IP telephony. The error here was recently pointed out to me by the CTO of a major equipment manufacturer, who noted: "Every building I go into has a big red fire extinguisher on the wall. Employees are trained to run for that fire extinguisher whenever there's a fire. Why not install a big red [public switched telephone network] phone in every room that's just for 911 calls - and not worry about providing it via IP telephony?"Well, duh. All these years we've been assuming that 911 had to be a built-in capability of the telephony system, which had to be able to provide the physical location of end users. That assumption worked fine when phones were tied to physical locations, but that's no longer the case. Instead of trying to retrofit newer technologies to outdated assumptions, why not require the physical location, such as the building or facility, to provide 911 services?The bottom line is always question assumptions - especially those about which you're surest.