Re: [apps-discuss] draft-farrell-decade-ni-02 - we think this is done...

[Carsten Bormann <[e-mail]>]

On Apr 6, 2012, at 12:28, Stephen Farrell wrote:

>> I believe for this to be the replacement of "unsecure links out of HTTPS", the secure media type issue must be solved in the base spec.
>> (Oh, and is there ever a need to discuss content-coding in this context???)
> 
> Not sure what you mean by "solved" can you elaborate?

The problem is that the ni: hash only secures the bytes of the payload.
(At least that's my assumption -- it might be worth saying explicitly what is hashed here, BTW.)

The entity I would have been retrieving over a secure HTTPS channel also has metadata.
Many of these are not critical, as they only pertain to the retrieval process (say, the Etag).
However, the media type is critical for correctly interpreting the returned result.
While I'm not aware of the proverbial contract text that means something different when interpreted as SJIS instead of UTF-8, attacks on the basis of swapping the media type are conceivable.  (Also, there is lots of potential for misconfiguration, damaging the reliability.)

I'm naming content-coding as another example that it is not quite clear what is the input to the hash function.
So maybe it is worth having another (normative) document that spells out how exactly ni:// is used in the Browser Web (among others, to solve the "unsecure links" problem)?

Grüße, Carsten