Draft. Draft — assembled from the working notes by `tools/build_book.py`. The source document is the authority; this page is a copy for reading, and it is regenerated from that source.

Status: Partially solved. The set is identified and read straight off the ROM — 33 glyphs, and they are block graphics, not line-drawing — and the “64 columns” question is answered by the video geometry. The glyph-by-glyph mapping to Unicode codepoints is deliberately not written yet: it has to be measured glyph by glyph, not asserted (see The mapping (open)).

The 2716 at UA15 holds 128 glyphs of 8×10, and the first 32 of them (0x00–0x1F) plus 0x7F are graphics characters rather than letters. They are the most stable thing in the ROM: the two char-gen variants differ in 93 of 128 glyphs, and the 0x00–0x1F set is byte-identical in both (docs/o1-rom-variants.md).

They are block graphics, not box-drawing

The natural assumption is the IBM/CP437 line-drawing set (─│┌┐└┘├┤┬┴┼). That assumption is wrong, and the manual says so in its own section title:

5.5 BLOCK GRAPHICS — Osborne 1 Technical Manual p. 25

Looking at them confirms it — the sheet is committed beside this file as o1-chargen-graphics.png, rendered by tools/romfont.py --glyph-sheet:

  • halves and quadrants (solid blocks filling half or a quarter of the cell)
  • staircases and triangles (the diagonal edges, drawn in steps because an 8×10 cell has no other way to do a diagonal)
  • a circle and a cross
  • 0x7F, a checkerboard/hatch fill — the glyph the site’s wordmark is hatched with

So the candidates to compare against are Block Elements (U+2580–U+259F) and Geometric Shapes, not the box-drawing block. That changes the mapping work from “guess a line junction” to “measure a fraction of the cell”, which is a much better position: a block glyph can be derived from its coverage.

What the set is for

It is what the machine used for graphics on a character display — the same purpose as CP437’s block and box characters, and the reason the O1 can draw frames and charts at all. Two consequences for this project:

  • A video personality (#24) has to render these correctly, not just the letters. docs/o1-memory-io-map.md §2a has the cell and attribute layout.
  • They are page furniture (#97). A period site’s rules, table borders, footer and bullets can be the machine’s own characters instead of CSS borders.

The mapping (open)

There is no codepoint table yet, and that is a method decision rather than a gap: each of the 33 glyphs is identified by looking at it beside its candidate and recording it with that evidence, not asserted from a description. The sheet is the evidence side of that work — tools/romfont.py --glyph-sheet renders it from the ROM, so it cannot drift from the ROM.

Next: render a Unicode Block Elements chart beside the sheet, identify each of the 33, then add the table to tools/romfont.py so the codepoints ship in the font.

Why 64 columns (see also §2a)

The recovered AUTOCPM art is 64 columns wide and the O1’s screen shows 52. That is not a contradiction, and the SCREEN-PAC guide supplies the missing piece: the machine separates logical width from physical width.

Logical screen size refers to the number of characters you can enter on a line without automatically inserting a carriage return. Logical screen size can range from 1 up to a maximum of 128 characters. … Physical screen size refers to the number of columns actually displayed on your screen. With SCREEN-PAC, this will be 52, 80, or 104. — Osborne 1 SCREEN-PAC User Guide, p. 13

A 64-column drawing is a logical line of 64 characters: it fits the 128-character video memory row, and the physical window slides over it. On stock hardware you see 52 of the 64 columns at once and scroll for the rest; with the SCREEN-PAC at 80 you see all of it.

The full statement, with the technical manual’s geometry and the attribute bits, is in docs/o1-memory-io-map.md §2a.

Sources