duplicity-talk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Duplicity-talk] Deleting large files from archive?


From: David Schneider-Joseph
Subject: Re: [Duplicity-talk] Deleting large files from archive?
Date: Tue, 29 May 2007 11:55:34 -0400

Kenneth and Charles, thanks for your thoughts. What do you think of Gabriel's idea? For a 1 GB file, would you be able to do a blind delete of 203 5 MB chunks, and possibly only have to edit 2 of the remaining 5 MB chunks?

On May 28, 2007, at 2:52 PM, Gabriel Ambuehl wrote:

On Monday 28 May 2007 20:26:55 David Schneider-Joseph wrote:

Presumably this is technically feasible, as the --remove-older-than
option manages to selectively delete only those files that are no
longer referenced, but it seems that duplicity does not currently
support this.  How would I go about doing this with the underlying
rdiff tools?


Technically it seems feasible to delete all tar chunks that contain ONLY THAT ONE file. So for a big file, this might work. However, you would have to somehow make clear to the restore process that the left over chunks in the first and last tar chunks (which almost always will be shared with other
files and thus cant be removed) are to be discarded for it not to get
confused.

As for implementation, I don't know enough about rdiff to comment on that.

On May 29, 2007, at 10:44 AM, Kenneth Loafman wrote:

David Schneider-Joseph wrote:
To follow up: I consider this to be a major missing feature. There's absolutely nothing to be done, and no future course for development? If you could point me in the right direction I'd be willing to work on a patch myself.

The problem is that to modify a file on the remote system requires that you download that file locally, modify it, then send it back. Since this is a multi-file tar archive, you would need to download the entire archive, decrypt, uncompress, delete the file, compress, encrypt, then upload the modified archive, making sure to delete the unused portions.

The remote system is *just* storage. It has no processing capabilities w.r.t. the archive itself.

...Ken

On May 29, 2007, at 10:54 AM, Charles Duffy wrote:

David Schneider-Joseph wrote:
To follow up: I consider this to be a major missing feature. There's absolutely nothing to be done, and no future course for development? If you could point me in the right direction I'd be willing to work on a patch myself.
Duplicity's design pretty much precludes it; consequently, a storage format change would be necessary to support this feature, and storage format changes are... a significant undertaking.

rdiff-backup also doesn't have this feature, though its design is much more amenable to the extension (and history editing is a much- requested thing there as well) -- but if you need duplicity's security guarantees, the lack of a straightforward mechanism for implementing reasonable/efficient history editing (without a complete disk format rework -- which would need to be done /very/ carefully to maintain duplicity's present security guarantees) is the consequence.


I can think of a few ways to modify the disk format to maintain reasonable security guarantees and place an upper bound on the amount of information that needs to be transmitted back and forth during the history-editing process -- but anything that comes to mind would require larger code changes than are presently feasible, never mind the complete lack of backwards compatibility.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]