Did not know of this when I wrote S3 support. Ability to resume large uploads would be good. Also allows supporting files > 5 gb, a S3 limit I was not aware of. NB: It would work just as well to split the object and upload the N parts to S3, but not bother with S3's paperwork to rejoin them into one object. Only reasons not to do that are a) backwards compatability with the existing S3 remote and b) this would not allow accessing the content in S3 w/o using git-annex, which could be useful in some scenarios. --[[Joey]]