| 26 Nov 2025 |
 hexa (signing key rotation when) | mail delivery has been restored | 11:51:24 |
 ma27 | Redacted or Malformed Event | 12:20:06 |
 somasis | yay :) thanks y'all | 16:22:45 |
 Arian | I was wondering if it would make sense to add a rewrite rule to Fastly for narinfos to drop any information about compression and push this into Content-Encoding instead. It would allow us to transparantly recompress the cache to new compression algorithms and levels . | 16:54:52 |
| 27 Nov 2025 |
 flokli | That's what snix nar-bridge does. The challenge with c.n.o is that it uses `nar/$fileHash.nar[.$compression_suffix]`, instead of `nar/$fileHash.nar[.$compression_suffix]`, so you can't easily find a NAR by it's nar hash without another lookup table somewhere. | 10:35:44 |
| Arian | oooh $fileHash is the hash of the compressed file? that’s a choice | 10:36:29 |
| Arian | T_T | 10:36:34 |
| flokli | * That's what snix nar-bridge does. The challenge with c.n.o is that it uses `nar/$fileHash.nar[.$compression_suffix]`, instead of `nar/$narHash.nar[.$compression_suffix]`, so you can't easily find a NAR by it's nar hash without another lookup table somewhere. | 10:36:36 |
| flokli | Yes | 10:36:39 |
| flokli | It's a good way to ensure we don't get too bored and run out of challenges. | 10:37:37 |
| flokli | A more sane way would be to have hydra push to some http daemon that does do all the rewriting, and uploads it to the right places. Or patch Nix to use $narHash. The URL in FileHash really doesn't matter. | 10:39:15 |
| flokli | In nar-bridge, we emit a path that encodes the castore root node as the path to the NAR, giving a subtle cue to a more castore-aware substituter, who can then potentially skip downloading most of the file entirely: https://git.snix.dev/snix/snix/src/commit/9c023a073e1b7298d322c26331331ccf5249cf0f/web/content/docs/components/store/snix_flavoured_binary_cache_protocol.md | 10:42:05 |
| flokli | (and use content-encoding to emit a zstd-compressed version, if someone wants the entire NAR) | 10:42:34 |
 edef | It would bump S3 bandwidth a bit, we'd have to do the math | 10:56:02 |
| hexa (signing key rotation when) | rebooting mimas/haumea for 25.11 update | 14:32:25 |
