)]}'
{"specs/in_progress/object_undelete.rst":[{"author":{"_account_id":597,"name":"Pete Zaitcev","email":"zaitcev@kotori.zaitcev.us","username":"zaitcev"},"change_message_id":"6ad1bcf9c3741da66337ed12dbd930a83e4ba736","unresolved":false,"context_lines":[{"line_number":11,"context_line":"Object undelete"},{"line_number":12,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":13,"context_line":""},{"line_number":14,"context_line":"This document describes options and modifications to Swift to prevent accidental"},{"line_number":15,"context_line":"object deletion, for example due to abused credentials or buggy applications."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"Problem Description"}],"source_content_type":"text/x-rst","patch_set":4,"id":"1a4dcd0f_bc838655","line":14,"updated":"2015-08-11 21:55:32.000000000","message":"Sorry for the wordsmithing right off the bat, but I thought the undelete helped you recover from deletions, not to prevent them. Maybe preventing the heaviest consequences of such deletions.","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"49ebe1481e5156b171ee5485bef84dee72e873a9","unresolved":false,"context_lines":[{"line_number":36,"context_line":"itself (container O), another one (container S) the symlinks to the objects. If"},{"line_number":37,"context_line":"a user uploads a new or modified object to container S the payload will be"},{"line_number":38,"context_line":"stored with a modified prefix and suffix in container O, and a symlink will be"},{"line_number":39,"context_line":"created in container S."},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"If a user deletes the symlink in container S the object itself is not deleted"},{"line_number":42,"context_line":"automatically in container O."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3afb71cf_c3a19495","line":39,"updated":"2015-06-24 22:09:18.000000000","message":"So will the option to enable versioning for a container only be available on container creation? It seems like migrating a container that already has objects would be time-consuming, and until it\u0027s done it would be difficult to determine which objects are currently protected.","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"49ebe1481e5156b171ee5485bef84dee72e873a9","unresolved":false,"context_lines":[{"line_number":39,"context_line":"created in container S."},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"If a user deletes the symlink in container S the object itself is not deleted"},{"line_number":42,"context_line":"automatically in container O."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"If an object is uploaded again (newer payload) the payload will be stored with a"},{"line_number":45,"context_line":"different timestamp suffix in container O (thus not overwriting any data), and"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3afb71cf_861a8ac4","line":42,"updated":"2015-06-24 22:09:18.000000000","message":"How will this interact with object expiration?","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"49ebe1481e5156b171ee5485bef84dee72e873a9","unresolved":false,"context_lines":[{"line_number":41,"context_line":"If a user deletes the symlink in container S the object itself is not deleted"},{"line_number":42,"context_line":"automatically in container O."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"If an object is uploaded again (newer payload) the payload will be stored with a"},{"line_number":45,"context_line":"different timestamp suffix in container O (thus not overwriting any data), and"},{"line_number":46,"context_line":"the symlink in container S will be updated pointing to the newer version."},{"line_number":47,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3afb71cf_c394f48b","line":44,"updated":"2015-06-24 22:09:18.000000000","message":"What should happen if an object is uploaded that has the same content as the object it\u0027s replacing? Will it be left to the client to to use something like --skip-identical, or could we do some de-duping server-side?","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"49ebe1481e5156b171ee5485bef84dee72e873a9","unresolved":false,"context_lines":[{"line_number":48,"context_line":"By using this approach all versions are kept, and the user has also the ability"},{"line_number":49,"context_line":"to get a list of \"deleted\" objects by listing container O."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Object naming in container O should contain the original object name as well as"},{"line_number":52,"context_line":"a timestamp suffix; there should be no leading prefix (like done in the"},{"line_number":53,"context_line":"versioning middleware). A small random suffix after the timestamp might be added"},{"line_number":54,"context_line":"to prevent collisions if objects are uploaded at the exact same time."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3afb71cf_e028069c","line":51,"updated":"2015-06-24 22:09:18.000000000","message":"Are we concerned about exceeding object-name length limits?","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"49ebe1481e5156b171ee5485bef84dee72e873a9","unresolved":false,"context_lines":[{"line_number":51,"context_line":"Object naming in container O should contain the original object name as well as"},{"line_number":52,"context_line":"a timestamp suffix; there should be no leading prefix (like done in the"},{"line_number":53,"context_line":"versioning middleware). A small random suffix after the timestamp might be added"},{"line_number":54,"context_line":"to prevent collisions if objects are uploaded at the exact same time."},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"For example::"},{"line_number":57,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3afb71cf_e34250ad","line":54,"updated":"2015-06-24 22:09:18.000000000","message":"Would the vector timestamps from Id85c960b126ec919a481dc62469bf172b7fb8549 be useful here?","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":597,"name":"Pete Zaitcev","email":"zaitcev@kotori.zaitcev.us","username":"zaitcev"},"change_message_id":"6ad1bcf9c3741da66337ed12dbd930a83e4ba736","unresolved":false,"context_lines":[{"line_number":69,"context_line":"Another important thing is that this implementation doesn\u0027t use internal COPY"},{"line_number":70,"context_line":"requests if a newer object version is stored, thus avoiding large data"},{"line_number":71,"context_line":"movements."},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"Alternatives"},{"line_number":74,"context_line":"------------"},{"line_number":75,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"1a4dcd0f_fc034e0e","line":72,"updated":"2015-08-11 21:55:32.000000000","message":"I think this needs a built-in expiration. Otherwise we\u0027re making users to roll their own inevitably.","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":597,"name":"Pete Zaitcev","email":"zaitcev@kotori.zaitcev.us","username":"zaitcev"},"change_message_id":"6ad1bcf9c3741da66337ed12dbd930a83e4ba736","unresolved":false,"context_lines":[{"line_number":83,"context_line":"these files periodically (for example after 24 hours)."},{"line_number":84,"context_line":"A possible drawback is that a rebalance might move the partition with the"},{"line_number":85,"context_line":"actual live data to another disk or node, and undeleting the data becomes much"},{"line_number":86,"context_line":"more difficult (because it can\u0027t be found easily)."},{"line_number":87,"context_line":""},{"line_number":88,"context_line":"Option 2"},{"line_number":89,"context_line":"--------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"1a4dcd0f_5faef4e6","line":86,"updated":"2015-08-11 21:55:32.000000000","message":"Even without a rebalance you need a special \"bypass\" API to fetch those. Relying on ssh/scp just isn\u0027t going to fly.","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"},{"author":{"_account_id":597,"name":"Pete Zaitcev","email":"zaitcev@kotori.zaitcev.us","username":"zaitcev"},"change_message_id":"6ad1bcf9c3741da66337ed12dbd930a83e4ba736","unresolved":false,"context_lines":[{"line_number":98,"context_line":"data."},{"line_number":99,"context_line":"This also ensures that deleted data is still replicated until the data is"},{"line_number":100,"context_line":"removed finally, and also ensures that data is always stored on the correct"},{"line_number":101,"context_line":"node, making it easy to actually retrieve the data."},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"Option 3"},{"line_number":104,"context_line":"--------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"1a4dcd0f_9f11dc06","line":101,"updated":"2015-08-11 21:55:32.000000000","message":"Same contra - needs an extra API, like a header \"Undelete: true\" or such. Still, does not seem too bad.","commit_id":"d6731108aef2a3ebb7c4518fd486375372f7efb5"}]}
