| 21 May 2024 |
 delroth | ah because the virtual method is upsertFile not uploadFile (the S3 backend has both methods, but the latter is just an internal helper) | 19:40:20 |
 puck | delroth also you missed that 272 suggests just vendoring the part of the aws sdk that does profile handling; rather than nuking it as such; iirc artemist was one of the people using the credential helper stuff from the aws sdk | 19:43:33 |
| puck | which is a bit more of an overkill solution, but | 19:44:09 |
| delroth | I think for the same reason that reimplementing it would be tricky, I expect vendoring it to require a lot of detangling | 19:44:22 |
| delroth | there's likely a bunch of internal deps on util libs for stuff like API queries to IMDS | 19:44:37 |
| puck | yeah, exactly | 19:45:05 |
| delroth | similar to why switching to the credentials loading part of that Rust lib would probably be a bad idea - it depends on Hyper | 19:45:06 |
| delroth | also: grep -r s3: tests/ returns no results 🤡 | 19:45:37 |
| delroth | so I added an item for that | 19:45:44 |
| delroth | I'm sure there's a minimalistic enough S3 implementation we can use to test stuff against | 19:46:54 |
| delroth | (at least in integration tests) | 19:47:00 |
 K900 | minio? | 19:47:03 |
| puck | i mean, generating the needed HMAC signatures isn't hard | 19:47:23 |
| delroth | I said minimalistic, not 100MB binary :p | 19:47:24 |
 raitobezarius | garage? | 19:51:51 |
| raitobezarius | * garage? | 19:51:57 |
 @irenes:matrix.org | ooo | 19:53:10 |
| @irenes:matrix.org | this is an exciting proposal | 19:53:13 |
| @irenes:matrix.org | also imo yeah giving the private beta people pad access is appropriate | 19:53:30 |
| @irenes:matrix.org | (noting that all y'all wound up in the private beta because somebody knows who you are and trusts your intentions) | 19:54:07 |
| @irenes:matrix.org | hmm where should I put comments on this design......... | 19:58:03 |
| delroth | design docs as markdown sent to gerrit when? | 19:58:25 |
| @irenes:matrix.org | I do love that thought | 19:59:29 |
| @irenes:matrix.org | just needs a repo for it really | 19:59:34 |
| raitobezarius | .oO(https://adr.github.io/) | 19:59:43 |
| @irenes:matrix.org | (good to know about) | 20:00:02 |
| raitobezarius | (example: https://spectrum-os.org/doc/decisions/) | 20:00:26 |
| @irenes:matrix.org | so briefly, I like and support this proposal | 20:02:42 |
| @irenes:matrix.org | I am a little worried about proliferation of features that lead to extremely long URLs - I also don't like how flake sources put a lot of options in the URL, but at least in that situation you can write most of them out as a nix clause instead | 20:03:24 |
| @irenes:matrix.org | I think it might be nice, perhaps as a separate effort afterwards, to review all the contexts where s3 URLs are likely to come up and make sure there's also a way to specify the things that are currently URL options outside the URL | 20:04:18 |
