The exact page text, read by your browser's voice.
The Internet Archive: Chapman (two copies), Howard, the atlas, the census volumes, the Polk volumes, the yearbooks, the patent lists, the Honor List
What it is: most of this build's nineteenth and early-twentieth century was read through one door. The archive.org copies actually used are named in their entries: the two Chapman items, Howard, the atlas (illustratedhisto00higg), the 1921 bound Polk volume (polksouthbendind00polk) and the 1956 directory, the census volumes, the Miskodeed and Madrigal yearbooks from the public library's copies, Fotia's study in full text, the patent lists and indexes, and the Honor List's images item.
Its standing, from this build's experience: its texts are transcriptions of scans, and the build's rule — learned across legs and stated in Part six — is that the page image is the source and the machine text a convenience with known failure modes: column interleave in the Polk scans, the Honor List's total OCR failure, Chapman's destroyed township table, the 1909 manufactures table's scrambled columns, and, on the transcription host the Howard text rode in on, the host's own warning that its text was generated using artificial intelligence. Every load-bearing table in this build was read from images or tallied against its maker's totals for exactly this reason.
Where it fails: on the same shelf sit the restricted doors: Miller's bibliography, borrow-restricted in both scans (401), and Ayer 1913, access-restricted — the archive holds them and will not show them to this seat, and the apparatus prints that state rather than a citation dressed as a read.
Sources:
- mishawaka-full/ — the items used, the failure modes, and the restrictions, as documented leg by leg.