Re: [Json] Canonicalization

[Nico Williams <[e-mail]>]

On Tue, Feb 19, 2013 at 9:00 PM, Francis Galiegue <[e-mail]> wrote:
> On Wed, Feb 20, 2013 at 3:36 AM, Nico Williams <[e-mail]> wrote:
> [...]
>>
>> Yes and no.  If the verifier and the signer both have the same
>> document then no c14n is needed.  If the verifier must reconstruct the
>> signed document -as opposed to receiving it from the signer- then the
>> verifier must reconstruct exactly the signed document or the signature
>> will not verify.
>>
>
> There is one thing I don't get: in any case, what is transmitted over
> the network is just a stream of bytes. One end writes that stream, the
> other reads it.

No, in this one case the two ends construct some data.  A good example
would be channel bindings (RFCs 5056, 5929), except that mostly that
has no structure, so it's not really a good example after all, but it
illustrates the point.

> In order for the receiving end to interpret that data, should signing
> be used, it needs to verify that the _byte stream_, not its
> interpretation, is correct. That byte stream MAY be JSON. It may not
> be.

That's just it: in this case the data isn't transmitted, only the
signature.  There's many protocols that transmit signatures (or
hashes) but not necessarily contents.  E.g., rsync.  What if you had a
JSON-based synchronization protocol and you're sending file metadata,
only there's a lot of it (e.g., large ACLs), and you're trying to
avoid sending it, so you send file names and metadata hashes, and if
the receiver's don't match then you send the actual metadata?

Nico
--