Stupidity comes in two flavors.
Sometimes, we just don’t think. We forget anniversaries, miss our freeway exits, sleep past our bus-stop, call one of our kids by the dog’s name. Stupid.
It’s stupidity, but it’s OK, for two reasons. First, we’re all bozos on this particular bus. Second, we’re stupid because the smart thought never came near us to be chosen instead. We were sleeping with our eyes wide open.
Then there’s Microsoft Outlook. Why, oh why would you design a system that replicates all server data on my local hard drive, then uses expensive, slow, network bandwidth for every operation when the free, fast connection between my hard drive and the CPU is there for the asking? I could, of course, use the “Work Remote” option, only then Outlook can’t use the network link, only the modem connection. This is ActiveStupidity.
(Disclaimer: I’m picking on Outlook because I use it. For all I know, Notes and GroupWise are just as bad.)
Then there’s IS’s current love-affair with browser-based applications. I use one on a regular basis, through a 128Kbps WAN connection, and what would be a 1-minute job with a client/server GUI instead takes 10 minutes of mostly thumb-twiddling.
IS has fallen in love with fat network architectures. Fully buzzword-compliant software companies call browser-based user interfaces “thin-client” applications, but they aren’t, because a) browsers aren’t thin, and b) the thinness of the client doesn’t depend on where it executes, only on how well the business logic has been isolated from the presentation. I’d like everyone who misuses the term “thin client” to write this on the blackboard 50 times: “A ‘client’ is a software process that makes a request of another software process. A ‘server’ is a software process that receives a request from a client and satisfies it.”
Architectures that ignore the capabilities built into intelligent desktops in favor of servers at the wrong end of a WAN link may or may not be thin-client architectures. They are fat-network architectures every time, though, because they require faster (read “more expensive”) WAN connections and bigger (read “more expensive”) servers to achieve a user experience that’s a fraction as satisfying as the one you can achieve through a local GUI.
IS has fallen in love with fat network architectures because they’re easier to manage than n-tier client/server architectures. Never mind that an increasing fraction of the workforce is mobile, where bandwidth is limited to modem speeds and the laptop is tethered to a hotel telephone. Our Total Cost of Ownership has gone down! Break out the champagne.
A correspondent from a prominent software supplier noted for its advocacy of fat-network systems took me to task a while back for missing their elegant beauty. His company’s system, for example, downloads Java applications that supplement the browser’s limited functionality in tiny little 100KB chunks. “It runs over a 56Kbps modem!” he enthused.
Let’s see. Factoring in protocol overhead, a 56Kbps link delivers a little over 5.6KBps. That means each of those tiny little Java applications will download in only … 17 seconds. There’s a satisfying response time. Never mind the sub-second response time you’d get with software installed on the hard drive. To keep the delay under 2 seconds you need to connect at full T-1 speeds. Take a look at your WAN. How many 128Kbps links would you need to upgrade?
I don’t get it. Pundits keep writing about the ultimate in thin-clientness: Keeping all their private files on nice, public servers so they can access them anywhere through public terminals … along with every cyber-vandal and cracker on the Internet. That’s much better. Of course, they’d still have to haul their suitcases and briefcases around, but it would easily eliminate 10 percent of the total luggage weight.
Given a choice, I’d rather lug my laptop and rent clothes from the hotel. No more lost luggage, no more suitcases to drag through the airport … it would be a slice o’ heaven.
I could really get behind a “thin-luggage architecture.”