- From: John Boyer <jboyer@uwi.com>
- Date: Mon, 6 Dec 1999 09:01:59 -0800
- To: "Peter Lipp" <Peter.Lipp@iaik.at>, "Tim Berners-Lee" <timbl@w3.org>, "DSig Group" <w3c-ietf-xmldsig@w3.org>
Hi Peter, > Incorrect. What is covered by the digest is ABDE, but within ABDE is a > *signed* assertion that the only allowable difference between the document > and ABDE is the addition of C between B and D. Can you please > state why you think this is insecure? I don't think you can. I think this overcomplicates things and nobody, or not many, will understand that concept. This is as if we sign ABCDE and tell the user to ignore C. Doesn't make sense. Either we want to sign ABDE, then we should sign ABDE, and if we wnat to put C into the picture, why tell them that we don't want to sign it, but still do in some obscure way, indirectly maybe. <John> Actually, we likely want to sign a document containing ABDE but ensure that future persons can only make modifications of the form given by "C if between B and D". Note that C still contains sufficient variability to allow the work necessary to 'close' or complete the document. This could, for example, be the act of filling out the 'office use only' section of a form, or adding multiple signatures to the document, or the act of 'code signing', where the code is represented by markup. The scenarios document contains more details on these examples. As for whether document closure overcomplicates things, we are talking about the same group of people that intellectually grasp the fact that cryptographic security can result from things like modular exponentiation and modular multiplicative inverse via GCD as well as details like the cascade effect and and the possible need for a nonce with regard to digital document 'fingerprinting' with an algorithm like SHA-1. Whereas the purportedly 'obscure' signing methodology I've been proposing amounts to little more than using cryptography to secure a message M which happens to contain a self-referential statement of the form "This message must originate from a document of the form ABCDE where C has the following characteristics...". Since university computer science coursework usually covers the cryptographic material mentioned above *after* the basic issues of computability such as the Turing halting problem and Godel's incompleteness theorem (which arise from problems pertaining to self-reference), it seems likely that the cryptography material is as complicated or *more* so, and that anyone who buys into it will not have trouble grasping the use of cryptography to secure a message M that contains a self-referential statement. John Boyer Software Development Manager UWI.Com -- The Internet Forms Company </John> Peter
Received on Monday, 6 December 1999 12:03:28 UTC