Dear TypeDrawers, what’s the best practice with rsync and UFOs? I neither want to wait forever nor zip every single UFO … Any recommendations? Thanks! Christoph
What are the options of rsync that you are using? I would play around with some of the options:
--checksum I guess the checksums to find modified files in an UFO take a while, but maybe it results in a faster transfer than if files that just have a new timestamp, but are identical, are included in the transfer.
--compress Compressing the files for transfer may or may not be faster than uncompressed files. If each file is compressed on its own, compression may be slower than uncompressed transfer. Not sure how rsync does it.
--whole-file Transferring whole files may also be faster if there are a lot of small files like in a UFO.
Another idea: If you have rsync access to a server, maybe you have also a login via ssh. Then you could set up a git repository with ssh access. git is quite efficient in optimizing the transfer, and you get version control as a benefit.
Comments
--checksum
I guess the checksums to find modified files in an UFO take a while, but maybe it results in a faster transfer than if files that just have a new timestamp, but are identical, are included in the transfer.
--compress
Compressing the files for transfer may or may not be faster than uncompressed files. If each file is compressed on its own, compression may be slower than uncompressed transfer. Not sure how rsync does it.
--whole-file
Transferring whole files may also be faster if there are a lot of small files like in a UFO.
Another idea: If you have rsync access to a server, maybe you have also a login via ssh. Then you could set up a git repository with ssh access. git is quite efficient in optimizing the transfer, and you get version control as a benefit.
I’ll try the rsync options as you suggest (and will report), and also I’ll see if git is an option.
was my initial setting. (-a archive mode, -E copy extended attributes)
(-z compress)
did not make any difference.
(-W whole-file)
was a bit faster.
I finally ended up with
leaving away -E copy extended attributes,
which was more than 3 times faster. I guess I don’t need resource forks.
Thanks again!
Right now it’s about the initial transfer, and therefore I didn’t test the --checksum setting yet. More to come …