An array-based serialization assumes that JWKs will never have structured 
values - JSON objects or arrays - that, or that when that case arises for 
future JWKs we'll have to invent an extension to the serialization instructions 
at that time.  It's not inconceivable that a future key format might include 
tuples of values, for instance.  The beauty of the current algorithm, being 
JSON-based, is that it's fairly future-proof.

-----Original Message-----
From: ⌘ Matt Miller [mailto:[email protected]] 
Sent: Friday, January 23, 2015 1:50 PM
To: John Bradley; Mike Jones
Cc: Richard Barnes; Jim Schaad; [email protected]
Subject: Re: [jose] Working Group last call on draft-ietf-jose-jwk-thumbprint

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

On 1/23/15 2:26 PM, John Bradley wrote:
> Yes,  it allows for new key types to work with this spec in the future 
> without needing to redo the spec and worry about downgrade attacks.
> 
> It seemed obvious at the time to use the JWK format.   It is 
> interesting that some people specifically don’t want that.
> 

This document is calling for handcrafting JSON.  It's eschewing whatever 
existing support in order to have a canonical format[1].

My original proposal still included the kty, which (I assume) disambiguates 
heterogeneous keys today.  I think it exceedingly unlikely that two different 
keys of different types will have parameter names and values that will collide 
with others, but I accept that some find it worthy of concern.

What if the serialization were a JSON array consisting of an array of pairs -- 
the member name followed by the member value?  For instance:

[["crv","P-256"],["kty","EC"],["x","Ze2loSV3wrroKUN_4zhwGhCqo3Xhu1td4QjeQ5wIVR0"],["y","HlLtdXARY_f55A3fnzQbPcm6hgr34Mp8p-nuzQCE0Zw"]]

Or in a flattened format, which is probably about the same amount of coding 
work:

["crv","P-256","kty","EC","x","Ze2loSV3wrroKUN_4zhwGhCqo3Xhu1td4QjeQ5wIVR0","y","HlLtdXARY_f55A3fnzQbPcm6hgr34Mp8p-nuzQCE0Zw"]

Or maybe we seriously consider SPKI.


- --
- - m&m

Matt Miller < [email protected] >
Cisco Systems, Inc.

[1] And yes, I fully appreciate how little weight this argument carries if a 
certain working group did something about canonicalization.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.22 (Darwin)
Comment: GPGTools - https://gpgtools.org

iQEcBAEBCgAGBQJUwsIjAAoJEDWi+S0W7cO10roH/jvyu4X/nTBW/sYw94l2xa6L
kDd4ZuQwQOPoSI0IawcZ8PQOk/php3eumS1EqTR/kxMGNbfDrc72Si3Kb6s4dRot
ZPn9ol8fu9HjaQUlT5J/f1tjLjClISVYZ/fAdu+GbxM/4C2tsgnyzWLPfLVE6l5k
MSGyV5j0MTmY3rBX5oQEcLRZc48EibL+R+gka0Mi/zSFdOkrVPG5i8E/qV1HwO2O
pIXHpvj/vuyrbFSjWml1Y9yiC4bMTK5HePgVzZ3+Jj1dBibg+TDH9gQMoInEIIIS
yKbt74TmZ5DLPobfPcmvBSofdMCPGgNIQWiF1OBJZpzj14d/APgAj+ICv0C4Ycc=
=+KOE
-----END PGP SIGNATURE-----
_______________________________________________
jose mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/jose

Reply via email to