Article by John Derbyshire

This past month of obsessing over chads and dimples has been a trial for the candidates, an endurance test for political commentators ("What's left to say?" they moan to each other in the hallways) and a vexation for many citizens.  For us old mainframe-heads, though, it has been a feast of nostalgia.  Punched cards!  I thought I had seen my last punched card 15 years ago.  Then, suddenly, they were all over my TV news.  As Verlaine whimpered:  "Souvenir, souvenir, que me veux-tu?"

I have been involved with computers since my schooldays.  They have provided me with a reliable source of income all through my adult life.  The first computers I encountered were mainframe machines, secluded in custom-built rooms with $40,000 air-conditioning units, trained attendants, and false floors beneath which lurked staggering quantities of wire. 

There were no proper terminals attached to these monsters, only a teletype for the operator.  Data came out printed on green-lined sheets of paper.  It went in on punched cards. 

My first few months of computing were spent getting acquainted with those cards.  Each card had 12 rows (numbered 12, 11, then 0 through 9) and 80 columns.  By punching holes in a column using a mechanical hand punch, you caused that column to have a meaning.  I have punched 2,000-line COBOL progams on the hand punch, each line a card with 40 or 50 columns punched—close to a hundred thousand downward jabs on the key pad.  Mis-punches were no problem, and I can now reveal the secret for replacing a chad in its punched hole securely enough to get it through the reader.  You lay the card flat on the table, press a free chad into the hole, and scribble over it with a soft lead pencil.  The chad will then stay in place even after passing through the card reader.  I used to think I was the last person in the world that knew this trick, but after watching the shenanigans in Florida this past month, I am no longer so sure.

Having punched your program onto a stack of several hundred cards, you took them to the computer room and delivered them to an operator, praying that he would not drop them (a computer program only works if its instructions are executed in the right order).  The operator, in his own good time, would feed them into the card reader—"face down, 9 edge first", according to the instructions on the reader.

In the later 1970s remote terminals came in, and networks, and on-line development tools; in the 1980s the PC appeared and "client-server" systems came into vogue.  The 1990s, of course, were the years of the Internet revolution.  I went along with all this somewhat grudgingly, mastered the baroque new programming languages, learned data modeling, object-oriented code, HTML tags and the rest.  In my innermost heart, though, I scorned it all.  I was still a mainframe-head.  As Rudyard Kipling pointed out:  "We've only one virginity to lose / And where we lost it there our hearts will be."

It was thus with some glee that I watched the Florida vote count.  Technology-wise, it was a trip down memory lane.  You and I may spend our free time surfing the Web; in Florida they are punching holes in Hollerith cards and feeding them into card readers—face down, 9 edge first, one hopes.  The card readers in turn send the data on the cards to mainframe systems, to be processed by programs written, in some cases, back when I was whanging away at the hand punch.

Why are voting systems so antiquated?  Of Florida's 67 counties, 27 used punched-card systems, 37 use the only slightly more advanced "mark-sense" cards developed in the 1960s, 2 still have mechanical voting machines and one relies entirely on manual counting.  This is about par for the nation as a whole. 

A large part of the problem is inertia, what systems people call the "installed base".  When you have a large computer system in place, it is awfully troublesome and expensive to switch to a new one.  Old programmer's joke:  How was God able to create the world in just six days?  Answer:  No installed base. 

There is a monthly newsletter called The Bell dedicated to the promotion of voting by Internet.  In the spring of this year they surveyed Florida voting systems.  One typical response to their queries came from the Supervisor of Elections for Tampa’s Hillsborough County, which currently uses punched cards:  "[The Supervisor] favors a touch screen system ... but is not ready to proceed with a purchase recommendation.  Changing voting systems is a huge undertaking ... Would like to change voting systems in the next 5 to 6 years."

For many of these localities, the sheer expense of changing systems is the main issue, and has actually incited rebellion.  Maryland state law, for example, requires counties to purchase computerized voting machines.  After estimating the cost to them as around $450,000, the commissioners of Dorchester County, in the rural eastern part of the state, have flatly declared their intention to break the law.  "I don't mind spending two or three nights in jail," remarked commission president Tom Flowers.  "They all like me out there."

And note that the height of ambition in Tampa is "a touch screen system".  Even their imagined future is out of date.  Touch screen—more properly known as "DRE", for "direct recording electronic"—systems are like those that turned up on some ATMs in the 1980s.  They are expensive to install and maintain, and produce no audit trail for later investigation.  They therefore increase the opportunities for vote-tampering by the people who program these systems, already a serious threat to electoral integrity. 

How difficult would it be for a programmer at one of the big voting-system vendors to corrupt an election?  Not difficult at all.  The program code is all "proprietary", which means the vendor need not submit it to independent audit—a principle that has been protected by the courts when tested.  The vendors frequently supply not only the software and hardware, but also the printed ballots, advice on voting procedures, maintenance and other services.  Vote tabulation logic is known only to the programmers.  Similar objections apply to any non-card computerized system, including one that employs the Internet.  With punched or marked cards there is at least some physical evidence of how the votes were cast.

Those Hollerith cards may thus be with us for some time yet.  A history of punched cards available on the Internet concludes with the following rather touching paragraph:

One of the last important uses for punched cards is likely to be voting.  Use of pre-scored punched card ballots was introduced in the late 1960s, and ... this format is, in 1998, the most widely-used computer-based election technology, although mark-sense ballots and direct-recording electronic voting machines are likely to replace punched cards in coming decades.

In coming decades!  So thirty or forty years from now, they may still be punching cards in eastern Maryland.  My entire working life will have begun and ended in the punched-card era!  I have no problem with this at all.  Just bury me face down, 9 edge first.