The External Codex

The constitutional material that landed this week — The Enrolled Text, the Amendment Atlas, and the glossary's new difference form — has a bibliography of exactly one item, and it isn't a science-fiction novel. It's my own PhD thesis: Formal Specification and Analysis of Danish and Irish Ballot Counting Algorithms, IT University of Copenhagen, 2013. Formal methods applied to voting systems. I should say so on the record, because the debt is structural, and because the way the site carries the debt turns out to be worth writing down.

Anyone who has read Schedule C alongside that thesis will recognise the relationship at once. Surplus transfers that conserve exactly what they transfer. Arithmetic in whole fractions, with the rounding deferred until there is nothing left to round. Ties broken by rule and never by discretion, so that the same ballots always produce the same seats on any desk, in any year. Proofs at every phase boundary that are not removed when the count runs in earnest, because they are the count. The loud halt in preference to the quiet error. None of that is invented colour. Those are the properties you end up caring about — sometimes because you proved them, more often because you watched a specification fail to have them — when you spend years formally specifying how Danish and Irish ballots actually become seats. The counterexamples were the real teachers. A vote-counting algorithm looks obviously correct right up until a model checker hands you the election where it isn't, and Schedule C's whole personality — its refusal to be clever, its preference for two plain rules performed exactly — is what that experience does to a person.

The same discipline runs through the Verification Doctrine and now through the enrolled text: the current wording as a derivation rather than an authority, amendments as differences small enough to be read completely, authority resting on anyone's ability to re-perform the derivation rather than on anyone's custody of a master copy. That is a formal-methods sensibility wearing constitutional clothes, and I got it from ballots before I gave it to the Concord.

The house rule I keep for material like this is already written down in the story bible: generalise, then fictionalise — the structure enters the work, the circumstances don't. This entry is the worked example from professional life rather than personal life. The Irish PR-STV count is not in the story. Denmark is not in the story. What crossed over is the shape: declared rules, conserved quantities, provenance on everything, a halt instead of an error — the exact list Sen files as one discipline in three uniforms, which was me noticing, somewhat late, that I had been dressing the same discipline in every uniform the setting could issue.

And that's where the title of this entry comes from. The site's Codex has one rule that makes it work: every entry is a named source's account, valid-for-its-author — their era, their access, their blind spots — never the record asserting fact. By that definition the thesis is a codex document too. It just sits outside the frame: a named source's account, complete with its author's era and access and blind spots, that the fiction draws on and can never cite. It has to stay external, for the same reason the record can't footnote it — an in-universe citation of a 2013 Copenhagen thesis would tear the frame the way nothing else could, and the Archive's voice is the one thing I don't spend. So it gets the treatment the site already has for things it points at without absorbing: referenced, not filed. The reference lives here, in the Journal, on the out-of-character shelf — which is, now that I look at it, the site keeping its own doctrine. Provenance on everything. Even on the doctrine.