October 8. The date I saw yesterday is today.
In the morning I opened the access record first. The one line I saw at half past five yesterday was as it was. The account is mine, the object of the query is the 1997 release register, the time is ten twelve a.m. In the date box it reads 8 October 99.
Yesterday that was tomorrow. Today it is today.
I looked at the line a long while and then looked for what had altered. Nothing had altered. Not one character has changed. What changed is the day.
At nine I ran the search. Two hundred and ninety-four.
Having written the log, I settled until ten what I would do.
That line records that I queried the 1997 release register at ten twelve a.m. I have not yet queried it.
There is an order I have learned in this room since January. When what is written and what has not been done stand against each other, I have chosen not doing. In the sixth month I did not count for a whole day, and in the seventh I did not run the search. Both times the result was the same. Undone, the record said done. What is done needs not a person but a record, and where a record has a box for a person, that box gets filled.
So today's work is two things. To query at twelve past ten and not to.
Querying, the record is right. Being right, I cannot tell whether that line was written first or whether I made it right. Not querying, the record is wrong, and twice already it has not been wrong.
I decided to try a third time.
At ten I set drawer seven on the desk and sat down. I pushed the chair back to where my arm cannot reach the terminal. I did nothing but draw a card, look at two digits, and write four.
At twelve past ten the card in my hand was 97. I wrote 1997. Four strokes.
It read consulted.
Until yesterday that box was empty. The time had not come, so empty was right. Today at twenty past ten there is a mark in it. It means someone opened it at twelve past ten.
The third time was the same as the others. Undone, it is written.
From eleven until twelve I pulled the whole access record. From the first of January until today. I pulled it once four months ago and there were some four hundred lines then. Today there are nine hundred and twelve.
I sorted it by date.
The top was not 99.
Two are 1 January 00 and two are 2 January.
The access record writes the year in two digits. It was so from January when I took it over, and no guidance to convert it has come yet. The access record is not in the twelve drawers. It is not an index but a record.
A 00 written in two digits is two years. It is 2000 and it is also 1900. The machine that sorts reads 00 as 1900, so it set the four at the top. The top is where the oldest line sits.
Then those four are queries from ninety-nine years ago.
That makes no sense. And to offer the side that makes sense, there must be grounds.
From one until three I looked for grounds. I set the four out and tried them one at a time.
I tried to tell them apart by the object queried. All four had an object. Two are indexes and two are catalogues. Both were programs handed over to me in January, so they did not exist in 1900. But the object box is a box that keeps the object's name, and the machine puts that name in as it stands now. There is nothing strange in an old line carrying a present-day name. It is not grounds.
I tried to tell them apart by the time. All four times are before dawn. Two fifty-one, two fifty-nine, three o'clock exactly, three oh four. This room is locked at six. That there is nobody in this room at three in the morning holds this year as it held ninety-nine years ago, so this is not grounds either.
Having written three in the morning down, I stopped once. That time is in the telegraph bundle in the box. It is the time when answers came on a frequency nobody calls. It is a thing of sixty years ago and has nothing to do with the records in use now, and only the time is the same. There were no grounds to write that it has nothing to do with it either, so I did not write it.
One came out. The four digits in the account box.
The archive code was changed last year. When the departments were reorganised in the winding-up after the merger, the four digits now in use came into being. Before that the number of places was different, I am told. In 1900 these four digits do not exist. So those four cannot be lines from 1900.
Having written it, I looked at what the line had done.
For eight months I have been writing down things that cannot be told apart. I wrote that I do not know which came first, and that there are no grounds for judgement. Today, for the first time, I found grounds and wrote it down one way.
Having told them apart, the side I told is worse.
If it is 1900 the line belongs to a time when this room did not exist, so it is somebody else's business. If it is 2000 it is a query I make with my account. What the four digits told apart is only that it is not somebody else's business. Before the telling it was one of two years; after the telling it is two days still to come.
For eight months there were places where not being able to tell was welcome. Not told apart, both stay alive, and with both alive the bad one is half. Today the half is gone.
At half past three I began writing what to send head office and stopped.
Among the five dates in the guidance there is 1 January 00. Yesterday's nine were explained by those five. Two of the four have the same month and day, so the same explanation can be tried. Only the test was run on the test machine, and the test machine leaves no access record. I confirmed that with my own hands yesterday and wrote it in the log.
I went to offer an explanation and was blocked by a line I wrote yesterday. What blocked it was yesterday's me.
At four the third-floor man came down and took a signature. On his way out he asked what I am working on these days. Writing years on cards, I said. He asked why one writes them.
Because a machine that sees only two digits cannot tell 2000 from 1900, I said. He asked how a person tells them apart, then.
A person looks at what comes before and after, I said. He said that will do, then, and went up.
After he went up I thought about what before and after are. Today's line has a before and an after. The before is my seeing it yesterday and the after is its becoming consulted today. Both are there, and still it will not be told whether that line was written first. The before and after one tells by is the order in which I saw it, and not the before and after held in the record.
From half past four until five I did cards. I made up two hundred.
While doing them I thought about writing dates in two digits. Two digits are a shortening to save a place. Saving it, two years come into one box. Come into one box, two years become one value. The figure that goes to tomorrow and the figure that goes to yesterday are the same value.
So a query dated tomorrow and a query from very long ago are the same characters in a record. It is for the same reason that yesterday's line became today without altering. What was written in advance and what was written after become the same characters, so from the record alone it cannot be told which.
At five I went down to the stacks. I meant to look once more at the box the 20 October line points to. It was in its place.
I came up and at ten past five opened the access record for the last time.
Today's lines were two.
The first is the one I saw in the morning. Ten twelve, 1997 Central Archive Release Register, consulted.
The second line's time was four minutes past five in the afternoon. At that time I was in the stacks.
There was a title in the object box. '1997 December Central Archive Release Register'.
The one I saw on 9 September ran from November to December. One that holds only December I have not seen before.
The result box read consulted.