Colophon
How this is set and made
A colophon is the note a printer leaves at the end of a book, recording who set it, in what type, and by what method.
The convention is worth keeping. A library that presents other people’s words owes its readers an account of the hands those words passed through on the way — which edition, which translator, which scan, and what was changed. That account is kept on each text, under its provenance note. This page records the rest: the type the site is set in, the reasoning behind the page itself, and the editorial method.
The type
Two faces, each doing one job. Both are freely licensed, self-hosted rather than loaded from a font service, and chosen because they belong to the same tradition as the texts.
IM Fell English
Titles and the wordmark
In 1686 John Fell, Bishop of Oxford and the man who built the University’s learned press, left it his collection of punches and matrices. He had bought them in Holland and Germany in the 1670s; the larger sizes were cut for him by his own punchcutter, Peter de Walpergen, and others came from the Dutch cutter Christoffel van Dijck. They were among the first old-style types cast in England, and Oxford has them still.
The digital revival is Igino Marini’s, made between 2000 and 2007. It reproduces the types as they were actually inked and impressed onto paper rather than as they were drawn — which is why the letters here look slightly worn, and why they are used for titles rather than for reading.
Junicode
Everything you read, and the labels above it
Peter Baker began Junicode in the late 1990s because medievalists had no type that could set what they needed to set — the thorns and yoghs and abbreviation marks of Old English and Latin manuscripts — and it remains the face that scholarship in this period is actually printed in. Its letterforms follow the Renaissance humanist romans of the late fifteenth century, which is roughly the tradition the earliest printed editions of these texts belonged to.
It is used here at a generous size on a deliberately narrow column, because the texts were written to be read slowly. It also carries real small capitals, which is why the section labels on this site are set in them rather than in the letterspaced capitals most websites use as a substitute. Small capitals are drawn at the height of the lowercase with matching stem weights, so a label sits inside the page rather than shouting above it — the difference between a typographic form and an imitation of one.
The page
The background is not white. Paper reflects the light around it and never reaches the brightness a screen is capable of, and a page of fourth-century prose does not want the glare of an empty document. The colour here is warm and slightly grey, closer to a book left open on a table than to a sheet fresh from a printer.
The reading column is narrower than the window allows. Long lines make the eye lose its place on the return, and the remedy is older than the web: set a measure, keep to it, and let the margins do nothing. Rules are hairlines in gold rather than borders in grey, which is the weight they were given when they were cut into a page rather than drawn on a screen.
Every colour on the site comes from a fixed palette of named values, and the build refuses to compile if a new one is introduced by hand. That sounds like a technical detail and is really an editorial one: it is how a page stays consistent when it is revised often.
The texts
Every author here died before 1900, and every translation used is out of copyright. That is not a formality — it is checked per text, against a named translator and a dated edition, before anything is published. Where the rule and the reality differ, the reality governs: a translation published before 1929 is in the public domain in the United States whatever became of its translator afterwards, and several texts here rest on exactly that.
Texts come from scans, from digital archives, and from the long-standing online collections that made this material available in the first place. Where a text was recovered from a scan, the scan is named. Where the quality of that scan affected the result, that is recorded on the text itself. Nothing is silently corrected.
Transcription and clean-up are done with the help of software, including language models, because a poor scan of an eighteenth-century octavo is exactly the kind of work that repays it. Nothing reaches the library on that basis alone: every text is read and released by hand, and the introductory notes are written by an editor rather than generated. The library grows slowly for that reason.
The build
The site is built with Next.js and served from Vercel; the texts live in a Postgres database. There is no advertising, no tracking of individuals, and no login. Analytics are limited to counting which texts are opened, which is the only question the library needs answered in order to know what to collect next.
The Stromata is edited by Keith Mason. There is more on the purpose of the library, and on how it came to exist, on the about page.
IM Fell types © Igino Marini, freely licensed. Junicode by Peter S. Baker, under the SIL Open Font License. Both are used unmodified except for subsetting to the characters this site needs.