The month was meant to be quiet, and on screen it was: the remaining application work – the piece picker, then the permission and invitation machinery for collaboration – was specified and mechanical, and moved slowly. The thinking went elsewhere: how a score should be cast off, and how the notes in a system should share out its width.

It started with Boxes & Glue: Ooloi's line breaker aligned with Knuth–Plass, down to TeX's own constants. Then Ulrik Bodén found Hegazy and Gourlay's 1987 report, 'Optimal Line Breaking in Music', out of print and nowhere online. Its springs let an accidental hold its room while the rest of the system gives way, which is how an engraver works and something Ooloi's design could not express. ADR-0037 now adopts them, and a new decision record, ADR-0055, specifies horizontal spacing: every rule reaches the breaker as data in one of four forms, and no correction pass is ever added afterwards. Vestigia Nulla Retrorsum has the reasoning; High Hrothgar says what it means – the method of the programs that lay out a whole piece at once, in a program that answers as you work.

None of it is built, but the specification has been checked. A model written from the ADRs alone, by a language model with no access to the code, cast off an 800-bar part in a third of a second, and its re-optimisation after an edit matched a full pass in every trial, down to the tie-break. The Seven Thousand Steps has the numbers, and what remains open.

Hammock Time is the one to read. It is about how this happened in a month I had set aside, and about something that changed: the model at the centre of Ooloi's layout came from a paper someone else found.

Next: the picker, then authenticated collaboration, shown in a video. That is the last of the plumbing. After it come spacing, beams, slurs and the page.

— Peter Bengtson ooloi.org