s3fs - I should also check this out, as it allows me to mount Amazon S3 on the server and access it from the outside via WebDAV, indirectly also to S3. For my pictures, it might be sensible in the long run to have two offsite backups with different techniques. For this purpose, I should also take a look at the s3fslight fork, as it is said to work better with rsync and that would be interesting for automatic backups. Both store files directly in S3, so they can be trivially downloaded. On the other hand, both have problems with the eventual consistency feature of S3, you have to run the synchronization multiple times, especially when you make many changes. Therefore, I should also take a look at s3ql, which forms a complete file system that only uses S3 as storage. This makes it more difficult to access file contents outside of s3ql - but there are also things like deduplication and encryption (in cloud storage, it seems more sympathetic to store things encrypted rather than unencrypted, even if you can trust some providers more than others).