- From: Takeshi Imamura <IMAMU@jp.ibm.com>
- Date: Mon, 29 Jul 2002 17:54:02 +0900
- To: merlin <merlin@baltimore.ie>
- Cc: xml-encryption@w3.org
Merlin, >You have a good point, and I've added text to the Limitations >part of the document that tries to describe this problem >and recommends placing the XPath/... transform first; if you >don't like this, please proposed text. > >However, you point out that placing the XPath first will >solve the problem in most cases. Adding the MAY text solves >only a miniscule number of additional cases: > >. Undoable super encryption must be performed >. AND this encryption must change the structure of the document, > preventing the XPath from going first >. AND the unprocessable encrypted data must occur in a part of > the super-encrypted document that is removed by the XPath. >. BUT the super encryption must not change the structure > of the document sufficiently to break the signature reference. > >Your MAY text means that in these few cases, the signature >MAY validate. I don't think that's good enough; either we >should state that the transform MUST fail or we should state >that the transform MUST proceed. In the latter case, we could >almost eliminate exception handling altogether. My intent was not to solve the problem but rather to enable to use the decryption transform in any places. I can easily think of cases where the decryption transform has to be performed first. For such cases, I don't think that we should suppose any place where the decryption transform is placed. >I'm not proposing the prohibition of full XPointers, I'm >proposing that we limit our super-encryption hack to just >handle barename XPointers. In the normal case, full XPointers >are processed fine. If, however, someone has an exception >URI that points into a super-encrypted block, then normal >XPointer processing cannot be performed and we have to >fall into a hack, which is to identify super-encrypted >EncryptedData elements based on their Id attribute and barename >XPointers. We're no longer doing XPointer evaluation, we're >doing a hack; I want to keep this hack simple, because the >added complexity of handling #xpointer(id('foo')) XPointers >doesn't buy us anything. > >Handling full XPointers in main-document exceptions is really >useful. Extending this into our hack isn't useful; it has >cost with absolutely no benefit. So I'm proposing to make it recommended or optional to implement. Considering that support of full XPointers in main-document exceptions is optional, it seems natural to make support of them in sub-document exceptions optional. However, I don't have so strong opinion on this, so I can take yours. Thanks, Takeshi IMAMURA Tokyo Research Laboratory IBM Research imamu@jp.ibm.com
Received on Monday, 29 July 2002 04:54:16 UTC