Re: comments on draft-abarth-mime-sniff-03

[Adam Barth <[e-mail]>]

On Mon, Jan 25, 2010 at 1:16 PM, Julian Reschke <[e-mail]> wrote:
> In general, defining certain behaviors normatively is a good thing, but it
> doesn't need to be all-or-nothing. Also, throwing around MUSTs frequently
> doesn't help in practice; what's more important is to *convince* the reader
> that doing something is the right thing. More inline.

Believe me, I'm doing a lot to convince the implementers directly.  In
fact, I'm going as far as sending them patches to align their behavior
with the spec.

> For instance, as Larry mentioned, (1) missing Content-Type, (2) mislabeled
> Content-Type and (3) invalid Content-Type (multiple headers) are distinct
> cases with different underlying reasons, so requiring the same behavior may
> not always be the best approach. So having different sections define this,
> and allowing people to pick the right pieces might be a good idea here.

As I wrote to Larry, having more permutations is antithetical to our
security and compatibility goals.

Adam