June 2. I decided to count.
Through May I tried not to. Counting is confirming, confirming has to be written up, and writing it up becomes a judgement. When there is no ground for a judgement it is better to write only what was seen — and in May what I saw kept increasing.
The post on the network board was deleted. Confirmation came on May 10. Then on May 20 the same post went up on another board. A different person posted it, and the text said it was a copy they had saved earlier.
That much I had expected. What has gone into hands cannot be deleted, I wrote last month.
What I had not expected came after. By the end of May the same post was in three places, and the three files differed a little from each other.
On June 1 I spent half a day finding copies.
The method was simple. Search network boards, search university archives, search the internet. The last is possible because the computer room has a line. My search terms were the lighthouse log and a few tide words.
Seven turned up. Four days earlier the same terms had turned up five.
Three on networks, two in university archives, two on personal homepages. All descended from what went up on April 27. Some had been recompressed; some had been split by page. One personal homepage had typed the text out as body text. Not photographs but writing.
That one took longest. Being text and not images it turns up in searches, and turning up in searches it had many visitors. The person who typed it wrote in a preface that they had keyed it in character by character. All hundred and eighteen pages, they said.
I read character by character twice. That means copied out.
I downloaded all seven and compared them against our file.
This far is confirmation. Confirmation is my work.
I made a table of the results. Having made it I sat for a long time.
Our registered copy has eleven lines present only in files. The internal downloads have eleven. The first posted copy has twelve. May's reposts have twelve to fourteen. The two personal homepages have sixteen and seventeen.
More than the numbers differing, what caught was that they differ in one direction. The lines did not decrease. Not in any copy.
I wrote view counts beside them.
Our registered copy, forty-eight internal lookups. The internal downloads, a handful each. The first post, four thousand two hundred. The reposts, one to three thousand. The two homepages showed visitor counters at twelve and fourteen thousand.
The copies with more views had more lines. No exception in a sample of seven.
I rebuilt the table three times. Reordered, re-based, the same. The copies more people read have more lines.
I had not meant to include view counts. Counting lines was the job. I put views beside them to tell the copies apart, not to set them against each other. They got set against each other because two columns side by side make the eye do it.
From here it reads two ways.
One. Copies with more lines are more interesting, so more people read them. Cause runs from lines to views. Read that way there is nothing odd. A version with more in it gets read more.
Two. Copies more people read grew lines. Cause runs from views to lines. Read that way there is no way to explain it.
To separate the two, measure the same copy at two times. Count the lines now and count again in a month. If views rise in between and lines rise too, it is the second.
On June 2 I made that table. Today's line counts and today's view counts for the seven copies. To count again in a month.
Having written it my hand stopped.
Only then did I see what making this table is. It is measuring.
Measure and it deepens — I read that somewhere. The survey log. The passage where the same place is measured over and over and the figure grows in one direction. While scanning I read it as a record of measurement error.
What I was making had the same shape as that table. A table for measuring the same thing repeatedly and writing down what grows.
I covered the table not out of fear but because, having recognised the shape, going on looking at it seemed likely to make me recognise the next thing as well.
I covered it and did something else for a while. Handled two requests, wrote the replies. My hands moved as usual.
Before six I opened the table again. It had been covered less than two hours.
Covering it does not stop the copies from increasing. Whether I count or not the homepage visitor counters are rising, and if lines are growing while they rise, that has nothing to do with whether my table is open or shut.
Not counting only means not knowing.
Written down, this was not a thought I had first today. There is a similar passage in the predecessor's notebook. A sentence my eye went past while scanning, which meant nothing to me then. I could find it now.
I did not look for it. Looking is reading, and I had just made a table showing that when readers increase, lines increase.
Having written that, I saw how strange the sentence is. I was avoiding reading. A person who manages records avoiding reading records is not doing the work.
But what is there to do instead? Read and it grows. Do not read and you do not know. Count and you are measuring; do not count and it grows uncounted.
I left the plan to count again in a month as it was. June 2's figures are down, so July 2 will do. Decide not to count and today's work becomes nothing; decide to count and I will still be doing this in a month.
Choosing between the two was the only thing I could do today, and having chosen, it did not feel like a choice either. The table exists, and a table that exists asks for its second column when the month is up.
Going home I put the table in the drawer. The connection log and the sheet of twelve lines were already there. It came to three sheets.
Set side by side they are three sheets about different things. A machine that attaches every night, a sentence that runs on, lines that increase. I worked at not joining the three into one sentence, and I knew that working at it was already joining them.
I locked the drawer and left. That locking is no use I learned two months ago. The motion of locking stayed anyway. Much of what people do is like that.