I have been battling with an ongoing problem with Amazon S3 for sometime now.
It has to do with serving video or any other media content directly from S3 to the client. Currently we have two very simple, very standard requirements:
1. Stream the video directly from S3 to Client 2. Download the video direct from S3 to Client (forcing a download prompt)
Now if you upload a video directly to S3 then simply have a link on your page to the video. By default the video will stream down. This is normal behaviour but when you want to force a download prompt it gets a tad tricky. After some investigating and experimentation I found that you could set the Content-Disposition header at upload time.
So I set was able to set this header using Bucket Explorer and it worked as expected. Problem solved? Not quite. Because the header information exists as meta data of the asset you can’t have a single asset served up two different ways.
This is a documented limitation of S3 right now as outlined in this post.
The “solution” as it seems is to create a copy (which incurs a cost) of the object and add the correct headers as required.
What this means is you end up having the same file duplicated existing twice with unique keys. I guess you would standardise a naming convention e.g. file.mov, file-dl.mov
Admittedly the cost of copying objects in S3 is next to nothing 0.12cents / per 1000 but it adds complexity to your application as you would need to trigger some offline process to go through and create the copies setting the correct header.
The best solution would be to have a query string variable which sets the header on request. So you end up with a Url looking like this:
Really hoping Amazon consider adding this as a feature soon.