Hello everyone,
Hope it's ok to post this here. This is a spec and an implementation of an open archive format for long term preservation on tape. Basically like tar but with indexes, checksums, parity, and optional encryption built in.
The specs:
REM-OBJECT: the archival object format (pax tar based, self-describing)
REM-PARITY: on-cartridge layout with Reed-Solomon parity
REM-ENCRYPT: optional per-object encryption, hybrid post-quantum (ML-KEM-768 + X25519 via X-Wing)
Implementation is architected as a tape component designed to be included in a larger archiving system via API.
Code is Apache-2.0, specs are CC-BY-4.0, conformance test vectors are CC0. Repo: github.com/archivetechie/remanence. Website: archivetech.org.
Built over about two months using Claude, ChatGPT, and Kimi in an adversarial panel setup, mostly run against a QuadStor VTL with a handful of passes on a HPE MSL3040 library with LTO-9 drives. Not in production yet, I wanted to get early feedback on the project first.
If you've got an LTO library and some spare time, I'd appreciate anyone trying it out. I've tested it quite a bit. But it's AI developed, so obviously don't trust your data with it until it's proven.
Hope it's ok to post this here. This is a spec and an implementation of an open archive format for long term preservation on tape. Basically like tar but with indexes, checksums, parity, and optional encryption built in.
The specs:
REM-OBJECT: the archival object format (pax tar based, self-describing)
REM-PARITY: on-cartridge layout with Reed-Solomon parity
REM-ENCRYPT: optional per-object encryption, hybrid post-quantum (ML-KEM-768 + X25519 via X-Wing)
Implementation is architected as a tape component designed to be included in a larger archiving system via API.
Code is Apache-2.0, specs are CC-BY-4.0, conformance test vectors are CC0. Repo: github.com/archivetechie/remanence. Website: archivetech.org.
Built over about two months using Claude, ChatGPT, and Kimi in an adversarial panel setup, mostly run against a QuadStor VTL with a handful of passes on a HPE MSL3040 library with LTO-9 drives. Not in production yet, I wanted to get early feedback on the project first.
If you've got an LTO library and some spare time, I'd appreciate anyone trying it out. I've tested it quite a bit. But it's AI developed, so obviously don't trust your data with it until it's proven.