You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have my complete FOOBAR collection in filestore.
Now I want, in my disk to rename a folder containing 1000000 TB of data.
Just a (folder or file) rename.
How filestore handle this EFFICIENTLY?
If we talk about torrents, filestore is the way torrents work.
EVERY torrent client has a "Set data location".
Because data moved often. This is the case.
So, I propose
ipfs filestore mv [-r] old_path new_path
-r includes all children folders and files (implies that old_path is a path to a directory).
No checks for hashes (if someone runs this command implies that he knows what he does).
Others propose add --nocopy with the same content from another location to replace old location.
This, I believe is wrong because new content must be hashed again. No. Just rename underling files.
I see, file paths stored in LevelDB databases.
So, it is easy to implement an additional rename operation (I believe).
The text was updated successfully, but these errors were encountered:
It's a so good idea that someone already proposed it :D (in a sense it's good because indepent people comming up to the same solution probably mean this is a good solution)
see #3981 (comment)
Actually I don't like this, for me filestore in the current form is a pretty good hack, but a hack non the less. I'm more for #8201 (oh my own issue what a surprise) and #7557, reflink makes more sense to me and include way less way to silently mess up.
I have my complete FOOBAR collection in filestore.
Now I want, in my disk to rename a folder containing 1000000 TB of data.
Just a (folder or file) rename.
How filestore handle this EFFICIENTLY?
If we talk about torrents, filestore is the way torrents work.
EVERY torrent client has a "Set data location".
Because data moved often. This is the case.
So, I propose
ipfs filestore mv [-r] old_path new_path
-r includes all children folders and files (implies that old_path is a path to a directory).
No checks for hashes (if someone runs this command implies that he knows what he does).
Others propose add --nocopy with the same content from another location to replace old location.
This, I believe is wrong because new content must be hashed again. No. Just rename underling files.
I see, file paths stored in LevelDB databases.
So, it is easy to implement an additional rename operation (I believe).
The text was updated successfully, but these errors were encountered: