>>>>>We want to add to our application the possibility of making remote backups of zipped data across the internet, probably in space we would rent from Amazon.
>>>>>
>>>>>Does anyone have experience doing this?
>>>>
>>>>I usually develop things like that so I can have them evolved when time comes. For VFP, you can use some available zip library, combined with a FTP library, which is something you will probably have to use to communicate and upload on the Internet (I would assume most of them would allow the FTP protocol to be used in regards to this). Otherwise, they would be able to support you for uploading from https, for example, with samples of code they can provide for various environments.
>>>>
>>>>For me, I have adapted my .NET utility to use the WinZip Command Line utility to zip and a FTP class to upload.
>>>
>>>Hi Michel,
>>>
>>>Thanks for the quick response.
>>>
>>>We already create zip files of the data, so that is solved. The question really is about how to upload and download files to space rented from Amazon.
>>>
>>>Another key issue is how to so organize things so that no customer can download data belonging to another customer.
>>
>>There seem to be various solutions that act as a conduit to use Amazon as a backup destination. You can Google [backup to amazon].
>>
>>You may ask yourself, why use these services when I might be able to set up something manually myself? The answer is they have solved some of the problems for you such as encryption, block-level/incremental backup etc. One example at http://www.cloudberrylab.com/amazon-s3-cloud-desktop-backup.aspx claims to cost $30 per computer which seems reasonable.
>>
>>Yes, keeping customer data separate is a key issue. One way would be to set up separate accounts per customer. These days I imagine this doesn't cost any more than to aggregate them all into one account.
>
>Thanks for the references Al and Craig. I want to make it seamless from the point of view of the client. I'll look into them.

One thing to keep in mind: S3 upload speeds are very slow, regardless of your upload bandwidth. The general prescription is to upload to a micro EC2 instance (with e.g., an FTP server, and a cronjob to move files to S3), which will be fast for both operations. Of course if your incrementals are not large, that's no an issue. If they are, this seems to be the best solution for using S3 as a target.

Hank