A new work in rehearsal is not a fixed object. Bars are cut, keys move to suit a voice, an orchestration is thinned because the room is small, and a number is rewritten on the Tuesday. The technical problem is not making the changes. It is that the same music exists in fifteen places at once, and they diverge.
Count the copies before counting the changes
The composer's file. The conductor's score. Every instrumental part. The repetiteur's copy. Each singer's vocal score, marked personally. The stage manager's book. The sound department's cue list. The title script. The projection timings. Any click that was built.
A change made in the first of those is a change that has not yet happened in the other nine. Rehearsal stops when one player is on an old page, and the time lost is the time of everybody in the room.
One authority issues, everyone else applies
The discipline that works is unglamorous. One person, usually whoever is preparing the music, owns the master. Changes are issued from there as numbered revisions with a date, and nobody edits their own copy from a verbal instruction unless the instruction is also going into the master.
The failure mode this prevents is specific and common: a correction given verbally in a rehearsal gets written into six copies by six people who heard it slightly differently, and the master, which was not in the room, still says the old thing. Three weeks later nobody can establish what the piece is.
Make the revision visible without opening the page
Every page carries its revision identifier and date in the footer, so that two copies can be compared at a glance rather than by reading. Where paper is used, printing revised pages on a distinctive colour makes an out-of-date part visible across a pit, which is a genuinely effective low technology answer used widely in new musical development.
Insert pages take letters rather than displacing numbers, so that a new passage after page thirty-four is thirty-four A rather than making everything after it wrong.
Bar numbers are load-bearing infrastructure
This is the point most likely to be got wrong and most expensive to repair. Almost every other document in the production is keyed to bar numbers: the calling book, the title script, the click, the projection timings, lighting cue points, and every private mark a player has written.
Renumbering the bars invalidates all of them at once. A cut should therefore be marked rather than removed, leaving the numbering intact so that bar two hundred and twenty is still bar two hundred and twenty for everybody who refers to it. Conventional marking of an excised passage exists precisely so that the material can be skipped without the map changing.
Where renumbering is genuinely unavoidable, it is a production-wide event rather than a music department decision, and every dependent document has to be revised in the same pass.
Files do not solve it, they relocate it
A shared folder with the current version in it is better than a pile of paper only if the naming makes staleness obvious and if there is exactly one folder. Copies downloaded to individual devices are the digital equivalent of a marked-up part in a locker: they persist, and they do not know they are old.
The properties worth insisting on are the same either way. A single authoritative location, an identifier visible on the material itself, a list of what changed, and an explicit statement of which pages to replace.
Marks are a second layer, and they are personal
A part that has been played for three weeks carries bowings, breath marks, cues written in from the neighbouring desk and reminders of things that went wrong once. Replacing the page discards all of it.
This is why players resist reprints that could have been inserts, and why a revision that changes two bars on a page should say so rather than arriving as a fresh sheet. Where a full replacement is unavoidable, the courtesy of saying which system changed lets somebody transfer their marks in two minutes instead of reconstructing them from memory across a run.
A change has a cost, and it is not the composer's alone
Cutting eight bars takes a composer a minute. It takes the librarian an hour, the players a rehearsal to re-mark, the stage manager a pass through the book, the title operator a script revision, and the video department a re-render if anything was cut to timing.
Making that visible at the point of decision is not obstruction. A production that knows a change costs a morning can decide whether it is worth a morning, which is a decision it cannot make while the cost is invisible.
What we cannot verify
Practices differ between opera houses, commercial musical development and concert premieres, and the conventions described here are widespread rather than standardised; no standards body governs music preparation. Claims about particular notation or file management software come from its publishers. What is not in dispute is the dependency: documents keyed to bar numbers break when bar numbers move.
The short version
- The same music exists in a dozen copies, and a change happens in one of them.
- One authority owns the master; nobody edits privately from a verbal instruction.
- Revision identifiers on every page, and coloured paper for revised pages.
- Inserts take letters so that later page numbers stay correct.
- Renumbering bars invalidates the book, the titles, the click and every private mark.
- A change costs a morning across the company, and that should be visible when it is decided.