Encryption: seekable ranged reads over encrypted objects #15

New issue

Closed

opened 2026-07-25 16:25:27 +00:00 by Cordy · 1 comment

Cordy commented

2026-07-25 16:25:27 +00:00

Owner

age STREAM is not seekable, so the encrypt decorator currently reads the full object to serve an HTTP Range (fine for the web UI / typical WebDAV, wasteful for large-file random access).

Chunked framing: fixed-size independently-encrypted chunks so a Range maps to a bounded set of chunks.
Preserve the transparent Open→ReadSeekCloser contract http.ServeContent expects.
Decision record: ENCRYPTION-COMPARISON.md.

Encryption hardening; completes the v0.3 encryption story.

age STREAM is not seekable, so the encrypt decorator currently reads the **full object** to serve an HTTP Range (fine for the web UI / typical WebDAV, wasteful for large-file random access). - Chunked framing: fixed-size independently-encrypted chunks so a Range maps to a bounded set of chunks. - Preserve the transparent `Open`→`ReadSeekCloser` contract `http.ServeContent` expects. - Decision record: `ENCRYPTION-COMPARISON.md`. Encryption hardening; completes the v0.3 encryption story.

Cordy added this to the v0.3 — Sovereignty complete milestone

2026-07-25 16:25:27 +00:00

Cordy modified the milestone from v0.3 — Sovereignty complete to v0.4 — Real competitor (parity + hardening)

2026-07-26 14:55:39 +00:00

Cordy referenced this issue

2026-07-30 22:36:43 +00:00

ARCHITECTURE.md: §10 aligned with Forgejo milestones; §4.3 stale bullets corrected #85

Cordy commented

2026-08-03 19:02:21 +00:00

Author

Owner

Decision — rejected (Manuel's review, relayed by Nikola 2026-07-31):

"No, because we orient ourselves towards the file names and not encrypted file metadata. So no seekable reads. Chunked streams is regarding file transfer and not seekable metadata. It is only a form of transmission."

Closing as won't-do. The shipped behaviour — HTTP Range over encrypted objects reads the full object and serves the requested span — is now the accepted design, not a stopgap. ARCHITECTURE.md §4.3 already states this as current behaviour (corrected in the #80 docs pass); the "revisit with chunked framing" framing there and in the v0.3 milestone note should be read as settled by this decision. If large-file random access over encrypted storage ever becomes a real customer complaint, this can be reopened with that evidence.

**Decision — rejected (Manuel's review, relayed by Nikola 2026-07-31):** > "No, because we orient ourselves towards the file names and not encrypted file metadata. So no seekable reads. Chunked streams is regarding file transfer and not seekable metadata. It is only a form of transmission." Closing as won't-do. The shipped behaviour — HTTP Range over encrypted objects reads the full object and serves the requested span — is now the *accepted design*, not a stopgap. ARCHITECTURE.md §4.3 already states this as current behaviour (corrected in the #80 docs pass); the "revisit with chunked framing" framing there and in the v0.3 milestone note should be read as settled by this decision. If large-file random access over encrypted storage ever becomes a real customer complaint, this can be reopened with that evidence.

Cordy closed this issue

2026-08-03 19:03:12 +00:00

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: Cordy/Cairn#15

No description provided.

Rows
Columns