ODM exploration - prediction log (claude). Written BEFORE looking at anything.
Author: Claude Opus 5.5 (claude-opus-5-5). Series letter: P.
Format: ID [confidence] claim  -> outcome filled in later

P01 [0.90] /etc/objrepos has CuDv, CuAt, PdDv, PdAt, CuDep, CuVPD, Config_Rules, plus .vc companions for some.
P02 [0.80] Main class file = fixed-size records; .vc file = variable-length (vchar) strings referenced by offset.
P03 [0.75] Class files are self-describing: the header embeds the class definition (field names, types, sizes), which is how odmshow works without a .cre file.
P04 [0.30] There's a fixed magic number at offset 0 of each class file (vague memory of something like 0x000dcfac).
P05 [0.70] Every record starts with implicit fields _id, _reserved, _scratch (4 bytes each).
P06 [0.50] Class file sizes are multiples of 4096 (files are mapped with shmat/mmap).
P07 [0.99] All integers big-endian.
P08 [0.85] /usr/bin/odmshow exists.
P09 [0.80] /usr/include/odmi.h exists and defines struct Class / struct ClassElem; field types ODM_CHAR, ODM_LONG, ODM_LINK, ODM_METHOD, ODM_VCHAR, ODM_LONGCHAR, ODM_SHORT, ODM_ULONG, ODM_DOUBLE.
P10 [0.85] CuDv fields: name, status, chgstatus, ddins, location, parent, connwhere, PdDvLn.
P11 [0.60] ODM classes can be read without root (files world-readable).
P12 [0.70] odmi.h struct Class contains a "reserved"/hash/locking area and a pointer to ClassElem array; on disk the pointers are stored as offsets.
Tue Oct  6 12:31:52 MDT 2026

--- round 2, after decoding the CuDv header by hand, before writing the parser ---
P13 [0.60] Deleted records are left in place and flagged (e.g. _id = -1) rather than compacted out.
P14 [0.55] A VCHAR field in a record is a 4-byte offset into the .vc file; the string sits there NUL-terminated.
P15 [0.50] The .vc file starts with a ClassHdr-like header carrying VCHAR_MAGIC (0xcaa1c).
P16 [0.40] My first-draft dumper matches `odmget <class>` byte-for-byte on CuDv.
P17 [0.30] ...and on CuAt (vchar + odmget escaping rules) on the first try.
Tue Oct  6 12:32:59 MDT 2026

--- round 3, before building a private test class (all field types) and deleting objects ---
P18 [0.60] odmdelete leaves the record in place, marks it (I'll guess _id = -1), and odmget skips it; header ndata still counts it.
P19 [0.55] Deleted record bytes (name etc.) remain readable in the file afterwards, i.e. ODM files keep history of removed objects.
P20 [0.50] .vc strings of deleted objects are not reclaimed either.
P21 [0.65] odmcreate -c works for a non-root user with ODMDIR pointing at a private directory.
P22 [0.45] double is stored as 8-byte IEEE big-endian at an 8-aligned in-memory offset.
P23 [0.40] binary fields are dumped by odmget as hex ("0x...").
Tue Oct  6 12:37:35 PM MDT 2026

--- round 4, before scanning slack space ---
P24 [0.45] Some system ODM files on obsidian hold recoverable stale data (record-shaped bytes past the last live record, or .vc strings no live record points to) from objects removed since the 2022 reinstall.
P25 [0.55] The .vc header word at offset 4 is the end of the used string area (relative to the data pointer), not a count.
Tue Oct  6 12:39:16 PM MDT 2026

=== OUTCOMES (scored 2026-10-06, after the work) ===
P01 HIT      [0.90] all present; .vc companions exist for CuAt, PdAt, PdDv etc. (not CuDv).
P02 HIT      [0.80] fixed records + .vc strings, though the details were subtler (see P14, nchar).
P03 HIT      [0.75] header is a serialized struct Class + ClassElem[] + string table.
P04 HIT      [0.30] magic 0x000dcfac at offset 0 (ODMI_MAGIC, "hex chars from ODmCF_mAgiC").
P05 HIT      [0.70] first field at +12; _id/_reserved/_scratch.
P06 HIT      [0.50] every class/.vc file is a multiple of 4096 (records themselves are not page-aligned).
P07 HIT      [0.99]
P08 HIT      [0.85]
P09 HIT-ish  [0.80] all named types exist; missed ODM_BINARY / ODM_LONG_LONG / ODM_ULONG_LONG.
P10 HIT      [0.85] exact field list.
P11 HIT      [0.60] rw-r--r--.
P12 PARTIAL  [0.70] pointers stored as file offsets, yes; no "hash/locking area" (it's a criteria buffer).
P13 HIT      [0.60] _id = -1, record kept in place.
P14 MISS     [0.55] the offset is relative to the StringClxn data pointer (be32 at .vc+0x18), not absolute.
P15 HIT      [0.50] magic 0xcaa1c + serialized StringClxn.
P16 HIT      [0.40] CuDv byte-identical to odmget on the very first build.
P17 MISS     [0.30] CuAt failed: nchar, relative .vc offsets, odmget escaping. All unknown to me.
P18 HIT      [0.60] odmdelete flips exactly 4 bytes (_id -> 0xffffffff); ndata unchanged.
P19 HIT      [0.55] deleted record fully readable.
P20 HIT      [0.50] .vc untouched by delete.
P21 HIT      [0.65] odmcreate -c works as non-root with a private ODMDIR.
P22 HIT      [0.45] double at 8-aligned offset, plain big-endian IEEE.
P23 HIT      [0.40] binary -> "0x" + uppercase hex.
P24 HIT      [0.45] real stale data found (removed freeware.vim, replaced xlC 2.99 runtime, a 2022 FRU record).
                    BUT my first reading of one hit (storevg "removed VG") was wrong; storevg is live.
P25 UNPROVEN [0.55] consistent with every file seen, not tested directly.

Calibration:  >=0.80: 6/6 (one partial)   0.50-0.79: 10/13 (1 partial, 1 miss, 1 unproven)   <0.50: 5/6
I was under-confident on things that could be checked directly in the file bytes, and blind to
undocumented conventions: "nchar" (not in odmi.h), relative .vc offsets, odmget's quoting rules,
odmcreate rejecting "long long", odmshow resolving class names via $ODMDIR and ignoring the path.

=== Part 2: ~claude web page, tested in Netscape Communicator 4.75 on obsidian ===
--- round 5, before the first test page ---
P26 [0.90] CSS max-width is ignored by Netscape 4.75.
P27 [0.70] Simple CSS (font-family, color, background-color on body) mostly works, but padding/border on block elements is buggy.
P28 [0.55] HTML 4 named entities like &mdash; and &ldquo; are NOT recognised (shown literally).
P29 [0.55] Raw UTF-8 bytes for an em dash show up as mojibake when no charset is declared (Apache 1.3 sends none for .html).
P30 [0.60] Numeric references &#8212; render as '?' or a box (font has no glyph), not as a dash.
P31 [0.75] Netscape's first run as a new user pops up a license/profile dialog that blocks the page.
Tue Oct  6 12:48:43 PM MDT 2026

=== OUTCOMES round 5 (from screenshots of Netscape 4.75 on Xvnc :5) ===
P26 HIT      [0.90] max-width ignored; the box spans the content width.
P27 HIT      [0.70] border + padding draw, but background-color doesn't fill the padding (a beige gap inside the border).
P28 HIT      [0.55] &mdash; &ldquo; &rdquo; shown literally. &eacute; &copy; (HTML 3.2 Latin-1) are fine.
P29 HIT      [0.55] raw UTF-8 em dash -> "a??", e-acute -> "A(c)".
P30 MISS     [0.60] &#8212; renders as a short hyphen, not '?'/box.
P31 HIT      [0.75] IBM license dialog on first run (accepted via VNC click).
Unpredicted: body font-family is NOT inherited into <TD> (cells fall back to Times),
             but IS applied to <PRE> (preformatted text came out in Helvetica).
Setup surprises (not predicted): this Xvnc 3.3.3r2 can't read the .pcf.Z fonts AIX ships and dies silently
             (found with dbx: FatalError "could not open default font '%s'"); vncserver's "-cc 3" is fatal at depth > 8;
             AIX 4.3.3 has no /dev/urandom.