)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":7847,"name":"Alistair Coles","email":"alistairncoles@gmail.com","username":"acoles"},"change_message_id":"a460c51f473b6118f2a0a18b74a7f24d387d0fe8","unresolved":true,"context_lines":[{"line_number":9,"context_line":"The probe test illustrates some inconsistency issues with object"},{"line_number":10,"context_line":"versioning."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"object_versioning creates a fresh version of potentially the same"},{"line_number":13,"context_line":".data file every time it finds it in the user namespace, so when"},{"line_number":14,"context_line":"overwrites only partially succeed there can end up being multiple"},{"line_number":15,"context_line":"versions of the same .data file copied into the versions container."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Furthermore, object_versioning doesn\u0027t attempt to preserve the time"},{"line_number":18,"context_line":"precedence of PUTs and POSTs; in the case of partially successful"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"0a8ce091_5dc615e6","line":15,"range":{"start_line":12,"start_character":0,"end_line":15,"end_character":67},"updated":"2024-11-26 18:38:58.000000000","message":"this is bad news for MPU cleanup!\n\nWe\u0027d ideally like there to be only one user object pointing to and MPU, ever. It\u0027s not possible to guarantee that with versioning, but we\u0027d at least like to know that there can only be one object in the versions container that points to the MPU - and know deterministically what its name is so we can do a HEAD to check for its existence.\n\nFollow-on patch explores one solution for the multiple versions.","commit_id":"c612b034fce66c6411468dbfc280ed6f24f0de16"},{"author":{"_account_id":7847,"name":"Alistair Coles","email":"alistairncoles@gmail.com","username":"acoles"},"change_message_id":"a460c51f473b6118f2a0a18b74a7f24d387d0fe8","unresolved":true,"context_lines":[{"line_number":17,"context_line":"Furthermore, object_versioning doesn\u0027t attempt to preserve the time"},{"line_number":18,"context_line":"precedence of PUTs and POSTs; in the case of partially successful"},{"line_number":19,"context_line":"POSTs, versions can end up with either stale metadata or metadata that"},{"line_number":20,"context_line":"belongs to a later version."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Change-Id: Ie880e401170105272dbaccd617a1830fb928bd05"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"2069db4b_59322b3f","line":20,"updated":"2024-11-26 18:38:58.000000000","message":"I don\u0027t have any simple ideas for solving this. We\u0027d need to do the equivalent of get_ondisk_files in diskfile, i.e.:\n\n* write .data and .meta files to the versions container\n* When doing GET/HEAD for a version, list all the .data and .meta and apply any .meta that is newer than the required version .data but older than the next version .data","commit_id":"c612b034fce66c6411468dbfc280ed6f24f0de16"}]}