Cache content to prevent re-download

zatricky commented 11 years ago

Situation/Issue

Example: Content on one client (A) is accidentally deleted or moved out of the shared folder. All other clients (B, C, D) remove the corresponding files.

Upon putting the content back (from backups/etc), A re-uploads the files to the server, and B, C, and D re-download these same files, resulting in a waste of bandwidth and possibly lost productivity.

Mitigation

The clients keep a cache of recently-deleted content. When it is found that the files need to be "put back", the content can simply be copied/moved back to the correct location, saving a download/upload.

Caveats/Dependencies/Notes

This very likely depends on #110.
Where possible, "cached" content could simply be a hardlinked version of the original file, thereby not taking up any extra diskspace. With a CoW filesystem this could be a snapshot/reflinked version of the file.
The client should be configurable in so far as how far back to keep deleted content as well as how much diskspace can be "reserved" for deleted content.
The server could benefit from a similar optimisation. Please let me know if I should create a corresponding issue there.
In order for client A to potentially benefit from a cache in this scenario, ALL files would need to be stored in the cache. This would necessitate use of reflinked and/or hardlinked "copies" of all files. The benefit would be the facilitation of "restore" functionality in the OwnCloud server interface. As above, please let me know if I should create a corresponding issue there.

--- Want to back this issue? **[Post a bounty on it!](https://www.bountysource.com/issues/998730-cache-content-to-prevent-re-download?utm_campaign=plugin&utm_content=tracker%2F216457&utm_medium=issues&utm_source=github)** We accept bounties via [Bountysource](https://www.bountysource.com/?utm_campaign=plugin&utm_content=tracker%2F216457&utm_medium=issues&utm_source=github).

etiess commented 11 years ago

It would be a great feature, you're right.

But for me it's not a priority: the stability of the sync is the most important issue for me. Do you know if there's any update concerning MD5 hash sum (https://github.com/owncloud/mirall/issues/110)? The more I read about that (and the more I have issue with sync), the more I'm convinced that MD5 would be the best solution.

In my opinion, MD5 calculation on the server is negligible compared to bandwidth usage and computing time for re-download/upload.

DeepDiver1975 commented 3 weeks ago

edge case we will not support

owncloud / client