Re: [jose] Comments on draft-barnes-jose-spi-00

[Richard Barnes <[e-mail]>]

>
> 1.       ** What headers are required and which can  be implicit – for
> example can the algorithm fields be implicit in the SPI?
>

If "spi" is present, then everything else is optional.


> **
>
> **2.       **Are the integrity value computed across the fully populated
> header or the SPI header?
>
I would suggest using the SPI header, just because it's simpler, but it
doesn't really matter.  If you trust the out-of-band process, then the two
have equivalent security properties.  I don't really care, since we
shouldn't be computing integrity values over the header anyway (ISSUE-4).


> ****
>
> **3.       **Is there a way to forward a message from person A which
> knows the SPI values and person B which does not?
>
I assume by this you mean
1. C -> A -> B, where C and A share an SPI, and B does not.
2. A does not have to do any cryptographic computations

In this case, the answer is "No".  The answer would be "Yes" if there were
no header integrity check -- A could simply fill in the pre-negotiated
parameters.  But the fact that the header is integrity-protected means that
A cannot do this, so the "spi" remains in place.


****
>
> **4.       **What is the correct algorithm for determining the JWS vs JWE
> in the event that all of the algorithms are implicit****
>
> **
>
Unless you know the SPI, you cannot make that distinction.  But you can't
reliably distinguish JWS from JWE right now anyway.


> 5.       **What happens if you have implicit parameters and explicit
> parameters and they do not match?
>
We should specify this, but I don't have a strong feeling for which option
we should choose:
1. SPI wins; b/c more trusted
2. Explicit wins; use SPI as default; override when necessary
3. Throw an error; malformed object


> ****
>
> **6.       **Is there a recommended way to determine what the SPI
> parameters are going to be?  Does the application need to pre-parse the
> message to get the SPI value or is there a recommendation that some type of
> callback be included
>
I would expect an API of a JOSE library supporting SPI to provide an
interface where an app could specify bindings between an SPI and a set of
parameters.  That way, the pre-negotiation protocol could use that
interface to configure the JOSE library, and the JOSE-using application
could just use the normal JOSE interface to process objects.


> ****
>
> **7.       **Can you make things like the IV be implicit?  Thus agree on
> a starting value and an increment and compute the new IV for each new
> message
>
That would be up to the pre-negotiation mechanism.  In general, this
doesn't seem like a good idea; you could think of secure ways to do it, but
it might be fragile.


> ****
>
> **8.       **If you are requiring that the values be populated by the
> application – does this require that you have a canonical encoding of how
> those values are placed into the header for the purposes of the integrity
> check?
>
This sounds like a good reason to do the integrity check over the SPI
header.


--Richard




> ****
>
> ** **
>
> Jim****
>
> ** **
>
> _______________________________________________
> jose mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/jose
>
>