# Current baseline and one-time instance conversion

The owner's October 11, 2026 change replaces historical installation upgrades with a fresh current baseline. Historical acceptance receipts remain evidence for their recorded source; this change does not certify a release or reopen unrelated held work.

## Supported installation and backup contract

- PHP 8.4 remains the minimum; development evidence uses PHP 8.5.11.
- The sole initial database migration is `202610110001`. Its 246 tables preserve the previous current physical columns, indexes, foreign keys and 101 CHECK constraints. Creation follows foreign-key dependency order. Generic migration/update and interrupted-install recovery infrastructure remain.
- The schema fingerprint is `55384d49dbdb3bab5ec4153daca692fd5112611976290a6151aebfca174c17f0`.
- Configuration must declare schema 3. The application key comes from the protected instance secret; inline-key/configuration upgrade actions have been removed.
- Full BK2 backups must match the current baseline and fingerprint and include the complete 244-table backup catalog. Historical archive adapters and the DynoComments/Panel Comments importer are removed. Native comments, guestbooks and other import formats remain.
- Restore retains the destination host configuration and application key. It re-encrypts protected fields, clears credentials according to existing restore policies, validates current authority, and preserves journaling/recovery protections. Future product updates remain a separate supported mechanism.

## One-time conversion

The supplied `dynostatic-20261011-224831-46e6c6a5.dynostatic-backup` remains unchanged. Its SHA-256 is `0f9840ed5660e5575626e6ca04478c1c3a7fedf70521f225c43389db37b43d80`.

A frozen copy of the original current source restored the authenticated historical archive into an isolated bridge database. The resulting current physical data was transferred into a database created by the new baseline, without transferring historical migration or data-upgrade ledgers. The new application's ordinary backup writer produced the converted encrypted archive. Production archive validation was not relaxed and the archive header was not merely relabeled.

The converted archive SHA-256 is `f5f9d32b7966dad8cec99b9af48e52b4a91cd91919e6df7c6fa6a1d4e0e67edc`. It uses the original archive password. It contains all 763 original rows plus one restore audit row; the original contained no embedded files or comment rows. The conversion preserves accounts, IDs, relationships and history rather than importing website content selectively.

## Recorded development evidence

Evidence is retained under ignored `artifacts/tmp/fresh-baseline-20261011/` and private `artifacts/private/backup-conversion-20261011/`.

- PHP 8.5.11 / MariaDB 10.11.19: direct baseline install, all-table structural parity and exact CHECK-clause parity with the original current schema, and idempotent second migration passed.
- Fresh installation and exact retry with a custom table prefix passed. The converted backup restored through the ordinary restore path into an installation with a separately generated protected application key.
- The untouched historical archive and wrong-password attempts were rejected. Native recovery also passed an unrecorded committed CREATE, rejected an old migration ledger without mutation, and rejected a weakened CHECK before returning to the exact original constraint.
- Original-to-converted and original-to-fresh-restored archive reconciliation found zero missing rows and zero changed original fields. Encrypted fields were compared through decrypted in-memory keyed digests. The fresh restore added its own second audit row.
- Focused configuration/import/admin/update unit checks and changed-runtime static analysis have passing receipts. Initial fingerprint, constraint-name and isolated harness failures are retained alongside their diagnoses and corrected confirmations.

The retained test fixtures now install the complete baseline. Historical backfill and prior-head migration scenarios are retired; current authorization, financial custody and recovery assertions remain. A permanent real-engine test covers committed-DDL recovery, obsolete-ledger refusal without mutation, and weakened-CHECK refusal. The four consumers of the shared financial fixture passed selected real-engine cases (4 tests, 110 assertions). Changed unit contracts ran 368 cases; the two remaining obsolete expectations were corrected and passed focused confirmations. Changed PHP fixtures passed syntax and static checks, with the final documentation-type corrections checked separately.

The verification manifest retains its check IDs and maps retired schema/test paths to the current definition and baseline recovery coverage. Current baseline implementation files and the shared financial fixture are explicit dependencies. Its 16 inventory tests and 590 shell-command syntax checks passed. These replacements do not reuse or relabel historical migration receipts; final qualification must select evidence against the resulting source and dependencies.

The additional `MySqlRoundTripContractTest` did not complete. Its first run exposed fixture installations sharing schema-global foreign-key names; those fixture databases are now isolated. One corrected confirmation reached its 400-second deadline during bounded migration progress, before its repository/comment/protected-file round trip. That outcome is **timed out**, and the wider joined workflow remains unaccepted. It does not replace or invalidate the separately completed supplied-archive restore/data reconciliation. No automatic retry or release campaign was launched.

This is development evidence, not release certification. The supported PHP/database matrix, full frozen-candidate campaign and extracted distribution archive still require the separate final qualification workflow. The source remains provisional until its required completion gates pass. No live installation, provider activation, stable promotion or publication is authorized by these receipts.

## Verification accounting

The owner approved a new finite 3,600-second development verification allowance for this change. Existing debits remain unchanged in `artifacts/tmp/foundation-audit-refactor-20261010/local-resume-20261011/testing-budget.json`; the `fresh_baseline_20261011` entry records this allocation, commands, outcomes, worker reserves and remaining shares. Failed, corrected, skipped and unexecuted outcomes remain distinct. A worker restart or new check does not reset the allowance.

Closeout debit: 3,440.066 of 3,600 seconds, including conservative preparation/diagnosis reserves; 159.934 seconds remains unspent. The wider native round trip needs a separately scheduled finite verification task; full supported-platform and extracted-archive qualification remain outstanding.