October 6. In the morning I opened the copy first.
The two pages I saw last evening were as they were. Both October 5, both with the year in four digits. The figure on the first page is two hundred and eighty-eight and on the second two hundred and eighty-five.
Two hundred and eighty-five is the figure I counted yesterday morning. Two hundred and eighty-eight is, by the rule, today's.
For eight months the copy has always been a day ahead. It laid tomorrow's figure on today's date, and that was the copy's habit. Yesterday the copy put out two pages. One is by the habit and one is not. The one that is not wrote yesterday's figure on yesterday's date.
After eight months the copy landed on its own date.
At nine I ran the search. Two hundred and eighty-eight.
Having written it down, I could not tell whether this is a good thing.
For eight months I have lived holding on to the copy's being a day ahead. Being ahead, there was something to compare against. If the copy holds today's figure yesterday, I can count today and check it. Agreeing, it is the rule; not agreeing, there is something to ask.
A copy that lands on its own date has nothing to compare against. If yesterday's copy writes yesterday's figure, I counted yesterday and so I know it already. There is nothing new to learn. For the first time the copy is right, and for the first time it is of no use.
At ten a requisition came. The third floor asked for one box of first-half 1998 tidying material to be released.
It is work like any day's. Take out the box and enter it, take the signature and send it out. In nine months I have done it a hundred times or so.
At eleven I pulled the register. It is a habit, checking the morning's.
Today's release was two lines.
The reason could be seen.
The release-register program was corrected to four digits in August. The index is not yet corrected and about half remains. On entry, the register draws the material number from the index. The register writes the date in four digits, and what the index hands over is in two. The two dates being different values, the machine takes them for different days.
The same day is entered as two days. One is 6 October 1999 and one is 6 October 99. To a person's eye they are the same day.
The corrected side and the uncorrected side were saying different years about one box.
Until noon I sat with the two lines. Deleting one looks right. One box went out and there are two lines, so one of them is a release that did not happen. A release that did not happen, sitting in the register, makes the register false.
I could not delete it. The release register is the same kind of place as the entry record. It keeps who took what and when, so if it could be corrected there would be no sense in keeping it. The three lines of the year 1900 left by August's test remain for the same reason.
I telephoned head office. The person responsible said to write the case up and send it. It is a symptom that will disappear once the index side is fully converted by the end of November, and duplicate entries occurring until then should be kept in a separate list and marked up in one batch afterwards. Not deleted, he said.
I made one more list. In nine months I have made five lists in this room.
Counting the five, they are these. The catalogue made in January because nothing had been handed over. The propagation table made in June to record where the wrong line had spread. July's, recording the board posts and their reproductions. August's, gathering the materials with no year entered. Today's is the fifth.
All five were made to write down what diverges. I made them believing that writing divergence down lets one set it right later. Of the four, not one has been set right. A list does not reduce divergence; it only makes divergence countable.
From two to four I did the rest of the sixth drawer. My hands did that and my head kept hold of the morning.
Counting what has increased in this room over eight months, it is all of two kinds. Figures and paper. The view counts rose, board posts rose, index places rose, cards rose. Yesterday's thirteen are paper too. Figures and paper may increase, and what has happened in the world does not increase with them. Thirteen more cards, and the box that went out is the same.
What increased today is a release.
A release is an act. It is a box coming off a shelf and passing into a person's hands. Two lines in the register say that act happened twice. A thing done once is written down as a thing done twice.
At half past three I went to the shelf and checked. In the first-half 1998 section one box was missing. One.
A divergence between the holding and the record is the first.
I held that sentence for about an hour. To settle which side diverges one must believe one of them. Believe the shelf and the register is wrong; believe the register and a box has gone missing. I saw the shelf with my own eyes. And in eight months I have passed many places where what was written came before what I saw with my own eyes. Today too, the only ground for believing my eyes is my eyes.
For eight months everything that diverged did so between records. The copy and the log diverged, the board and the access record diverged, the paper register and the screen diverged. Which was authoritative could not be settled, and both sides were records. While records quarrelled, the boxes on the shelves were always their own number.
Today, for the first time, one of the sides has a box in it.
One box went out and the register says two. One might say the register is wrong, except the register was made uncorrectable, and the reason the company made it so is in order to hold the register to be right. What one is to do when the thing held to be right is wrong is not in the form.
At four the man from the third floor came down and said he had received the box all right. I said it was entered twice in the register. He laughed and said then give him another one. There isn't one, I said. Then what does twice matter, he said, and went up.
After he had gone I thought about it again. It does not matter. One box went and he received one and today's business ended there. Where it comes to matter is with whoever reads this register later. That person is not here today.
At half past four I wrote one line into the duplicate-entry list and closed it.
At five I pulled the register one last time. I meant to count today's again.
There were three lines.
First line, 6 October 1999. Second line, 6 October 99. There was a third line.
20 October 1999.
The material number was the same, the recipient the same, the quantity the same. It is the box that went out today.
October 20 has not come.
I went down to the stacks again. I looked at the first-half 1998 section. The one gap that was empty this morning was empty as before, and beside it was full.
Which box goes out on the twentieth is written in the register. That box is still on the shelf.