Re: Questions about draft-abarth-mime-sniff-00

On Mon, Apr 6, 2009 at 6:26 PM, "Martin J. Dürst"
<duerst@it.aoyama.ac.jp> wrote:
> For proxies and firewalls, the story may not be as simple.
> How will they be able to distinguish between content sent to
> browsers (where they can assume that the algorithm in this draft
> applies) and other content?

Their behavior will depend on their purpose.  Some proxies will wish
to be conservative and err on the side of safety (i.e., filter/block
assuming both sniffing and non-sniffing user agents).  Ultimately,
having two kinds of clients (sniffing and non-sniffing) is an
improvement on having the N types of clients we have today (where each
rolls its own sniffing algorithm).

Another option we could consider is supporting something like IE8's
X-Content-Type-Options: nosniff (whereby proxies and servers could
instruct their clients not to sniff).

Adam