Summary

  • On 12 November 2003, Runeberg changed its next-page link to seek pages without a whole-page completion mark.
  • Nine days later, its statistics adopted the same rule. Earlier corrections survived, while many works moved into lower tables.

A promotion in reverse

Finishing a book usually earns you permission to put it down. At Project Runeberg in November 2003, some books acquired the opposite privilege: another appearance on the work list.

The Nordic literature project had a practical problem hiding inside a pleasant word. “Proofread” could describe somebody checking an entire page. It could also describe somebody noticing one misspelling, fixing it and getting on with lunch. Both actions left useful text behind. Only one told the next volunteer how much of the page had received attention.

On 12 November, Lars Aronsson announced a change. The link for finding unfinished work would now search forwards from the page the reader was on, rather than starting at the beginning of the book. Its English label changed from “First not yet proofread page” to “Next not yet proofread page”.

There was a second, more consequential adjustment. The link would also visit pages that had already received corrections but lacked the checkbox marked “whole page OK”. That box now decided whether the system counted a page as proofread. Some pages had just lost their graduation certificate.

Let the reader reach the text

The setup made sense for a project trying to stretch a small editorial office across a scattered volunteer community. In his volunteer account dated January 2003, preserved on a page carrying a 2012 revision footer, Aronsson described delays when editors struggled to assist everyone. More self-service was the proposed relief.

Online proofreading, he wrote, had arrived in 2002. Runeberg put facsimile images and raw optical-character-recognition text on the web before proofreading. Beneath the image, a link invited readers to fix errors through a web form. A recent-changes page let volunteers follow one another’s efforts. The printed page, editable text and evidence of other people’s activity could all be reached without waiting for an editor to answer an email.

That opened the door to small contributions. A visitor who could repair one word had something worth contributing. The awkward part came later, when a record of intervention was asked to stand for a record of completion.

The November checkbox gave that second statement its own place. Saving a correction recorded the work; ticking the square declared its extent. Meanwhile, starting the search from the current page made “next” depend on where the volunteer already was. Navigation and completion were being made more specific together.

Nine days for the tables to catch up

Aronsson’s announcement asked for comments on the change and acknowledged that something might have been missed in the transition. He also suggested bringing the proofreading statistics into line. Works, chapters and pages previously considered finished would slip down a level.

On 21 November, Hans Persson reported that he had changed the statistics. Only pages marked as wholly checked now appeared as complete. Many works had consequently fallen into later tables.

All earlier corrections remained. Persson said fully marked pages would receive priority because, in his judgement, their average quality was higher. His explanation was pleasingly ordinary: an unmarked page might already have been checked throughout, or someone might merely have corrected a single spelling error.

The longer-looking work list therefore contained a question the old total had swallowed. How far had the previous reader actually gone?

Runeberg’s books had survived the change with their corrections intact. The completion figure had to endure a little embarrassment. It was finally being asked to say what it knew.