- From: Rich Salz <rsalz@datapower.com>
- Date: Thu, 17 Oct 2002 11:25:20 -0400
- To: reagle@w3.org
- Cc: XML Encryption <xml-encryption@w3.org>, ross@contivo.com
I found your note very interesting. Serendipitously, the infoset/byte "conflict" has just been coming to the forefront in SOAP/DSIG. I should clarify that while I have much experience here, I am not a member of the XML-Enc WG. It seems to me there are two major issues. The second one is easier to deal with: does it make sense to try validate a partially-encrypted document. I am strongly in favor of saying that this is not a requirement on XML-Enc. This issue should be addressed by adding text to the spec. > These two points together argue strongly that it is the contents of > the infoset that should be signed, and not the serialization of the > infoset. The second issue (the first one you bring up) is more complicated, but I believe that -- for now -- it has a similar simple answer. Cryptography is based on byte-streams. Interoperable cryptography requires everyone to have the same serialization mechanism. Fortunatately, XML *is defined as* a serialization mechanism. :) Until there is a standard way to render the infoset (or more accurately here, the PSVI), "encrypt the infoset" is just plain impossible. Not hard or infeasible. Impossible. If you accept that encryption (and signature, although to a lesser extent) implies and/or requires serialization, then I believe most of your issues and message text become moot. I recognize that this is not optimal from a schema viewpoint, nor as flexible, but the goddess of crypto must be fed bytes. At this point, I am beginning to think that infoset/serialization is an issue for the TAG to consider. > A possible approach to resolving this problem, which Schema would > encourage you to consider, is to specify not a specific element, but a > complex type of encrypted data. This would allow the schema author to > specify element X and an encrypted form of X as alternatives. Don't the EncryptedData, et al, types provide this? > children of the_corn are encrypted. Oi. > which in turn means that validation of documents > containing encrypted content is not practical for a processor that > does not have access to the decryption keys. Perhaps the spec should make it explicit, but I'd argue "well, duh!" :) /r$
Received on Thursday, 17 October 2002 11:28:19 UTC