D012 - What a package writer still needs, and what it does not

Writing a package means emitting fourteen entries. Reading was enough to establish which fourteen; it is not enough to fill them. Characterised by looking at what each holds - sizes, how much is printable, and whether it announces itself:

entry	what it is	how known
0x10, 0x20	the key material	used by the reader already
0x200	a filename table	92% printable, names the icon
0x1000	param.sfo	begins with the PSF magic; shadPS4 has a reader
0x1001	a playgo chunk	begins plgo; shadPS4 has a reader
0x1003	XML	UTF-8 BOM then <?xml version
0x1200	a PNG	the icon entry 0x200 names
0x409	zeros	8192 of them in the sample
0x1, 0x80, 0x100, 0x400, 0x401, 0x1002	unknown	sizes and entropy only

So the gap is narrower than "a package is opaque" and wider than "just write the bytes". Six entries have no established meaning, and two of those (0x400, 0x401) are high-entropy and plausibly digest tables - plausibly, which is not a thing to build on.

The lawful route is a writer, and there is not one here

Identifying a PNG by its magic is reading a filename. Working out what belongs in 0x400 by staring at 1024 bytes of entropy is reverse-engineering, and it produces exactly the convergence the provenance rule exists to prevent.

This is the position mkself was in before an open-source writer turned up in the emulator kit and supplied every block a reader never looks at. The same thing is needed here: a packaging tool whose source states what it puts in those entries. LibOrbisPkg is named by the extractor this project already cites, and is not in the local kit.

Recorded as a blocked item rather than a hard one. The distinction matters: more effort against the current sources does not close it, and the honest next step is finding a citable writer rather than trying harder.

Status: decided - writing is scoped and blocked on a source, not on work.