# Shovel this folder's contents, recursively, up to mybucket.
# If a file is already there, it is not sent again
s3-upload-missing . mybucketname .

# Do it with verbose output
s3-upload-missing --verbose . mybucketname .

# Specify a prefix in the destination bucket.
# A slash at the end is implied if not provided
s3-upload-missing --verbose . mybucketname uploads

# Allow the public to read the files (web-accessible)
s3-upload-missing --verbose . mybucketname uploads --acl=public-read

# If a file is unreadable, chmod it momentarily so we
# can read it, then send it to S3 with "private" permissions.
# Then chmod it back to 000
s3-upload-missing --verbose . mybucketname uploads --acl=public-read --chmod-if-needed

# Also remove any remote files that do not exist locally.
# Use with care
s3-upload-missing . mybucketname . --delete

You must populate ~/.aws/credentials with your key and secret, like this:

[default]

aws_access_key_id = xxx

aws_secret_access_key = yyyyyy

TODO: support command line arguments for these as well.

"Why not use s3cmd?" s3cmd works fine, but we have a peculiar need to successfully upload files with permissions 000 and give them the private acl on s3 (the --chmod-if-needed option). This is very useful when transitioning from local files to s3 with uploadfs. Also, s3-upload-missing may be faster.