I can't upvote you enough. Too few people realize that the server returned mimetype actually tells the browser how to handle the content.
It's what's intended, but it's not what's true. Usually this bites on video, but it even bites on, say, SVG:
http://stackoverflow.com/questions/10261460/internet-explore...
Ostensibly, this is to protect users. Quoting TechNet:
When files are served to the client, Internet Explorer uses the following pieces of information to decide how to handle the file:
1. File name extension, the corresponding ProgID and CLSID for the registered handler of that file name extension.
2. Content-Type from the HTTP header (MIME type), the corresponding ProgID, and CLSID for the registered handler of that Content or MIME type.
3. Content-Disposition from the HTTP header.
4. Results of the MIME sniff.
-- http://technet.microsoft.com/en-us/library/cc787872(v=ws.10)...
Yes, I'm aware that this is probably a workaround for broken web servers, or brain-dead server admins. This workaround should never have been deployed, as it breaks interactions with non-broken web servers.
You can't call either browser broken here. One is doing what the spec says is correct. The other is doing what the user says is correct.