Last month, I mentioned that I attempted to use Amazon's S3, through JungleDisk to store music. Now that Google is potentially coming out with their Platypus storage service, I was thinking about this a little more, and I think that I know what use cases I would need before I would really use a service like this.
When I copy a file to the "online drive", I will want to be able to access it in the following ways:
- Through standard HTTP 1.1 protocol. – If this is a music file, I should be able to access it directly from xmms or iTunes.
- Through a cached virtual file system – Access to the file will cause it to be downloaded locally, so subsequent accesses will be fast. Also, any changes made to the local copy, would need to be syncronized back to the online version
- Though a non-cached virtual file system – From the client's point of view, this will appear as a standard file system. But an open will cause the http 1.1 session to be opened. Every seek will cause an http resume at the correct offset. And every read will just pipe the data off the http socket to the caller.
It doesn't matter if these features are implemented in the core, or some third party developer implements them.