When the Computer Tells You To
Michael Ubaldi, July 4, 2007.
From six years of dilettante's work with computers, I have learned more about contingencies and options than hardware or software. Nothing ever goes right at first, and I am wiser for the trial.
Late last Thursday night, my combination modem-router died in an ataxic flutter of LEDs. This was the second model sold to me by SBC in four years, a replacement that cost $150, which — considering retail prices of these little transmission boxes and their presumed longevity — was about ten times too expensive. When the previous loss occurred, however, I hadn't asked SBC if its house brand were indispensable to connecting my computer to the internet, and SBC hadn't disclosed. Payment was justified, thinly, by convenience.
This time, telephoning SBC's acquirers, AT&T, I inquired about store-bought equipment. Could a third party modem and router be used? The answer was Yes, my sentiment was Good Riddance. I ordered the necessary items, good quality pieces for less than their predecessor. They arrived yesterday, and immediately after work I began building a new link to broadband.
As per several laws of project management, the activation — all of twenty minutes — gave me 90 percent of what I wanted. Extracting, from the modem and router, that remaining tenth of performance took me the rest of last night and, leisurely, from about nine in the morning to four-thirty this afternoon. Amid a lot of desultory flipping of switches there was some education, some of it simply interesting and one part directly applicable. "If you set MTU to 1492 bytes," someone had written to another in a corner of the web, "you will solve most router problems." What on earth is MTU? Maximum Transmission Unit, and I didn't know that five hours before now. My router had MTU set, by default, to 1500. I changed the parameter and all problems were solved.
What makes the difference? In an analogy of a man collecting items from a conveyor — should the rate of the belt be too high, the man can't work fast enough and, paradoxically, slows down production. But now, an assembly line brought down by a motor deviation of one-half of one percent?
Jaron Lanier, writing in Discover magazine, cites the limitations of traditional programming as "explicitly defined protocols, a very precise but rigid approach," wherein "flipping just one bit in a program might cause it to crash." Both the modem and the router identified AT&T's service and adjusted settings to allow my computer to access it, but the disparity of a single variable denied me full use of my product. Lanier muses that "approximate pattern recognition...can become very reliable at understanding a complex system." We enthusiasts can have our electronics, and tough lessons, on better terms. Can a less logical order be one that is more rational?
See more: Our Man Ubaldi