The discussion about "least astonishment" led me to review this document and I have to agree that it raises some issues of "astonishment." It seems to me that the draft unnecessarily joins two separate concepts: 1 - Specifying a portion of a document, and 2 - Providing an identity check on the complete document. The mechanism proposed for #1 seems to be well specified. I have some comments on #2. I think what the "hash sums" are trying to do is verify the "correct" version of the whole document. I can think of a number of such checks off of the top of my head: 1 - md5 hash 2 - length 3 - charset 4 - Content-Id 5 - timestamp 6 - Content-Language Obviously not all of these attributes would be available from every source, but some of them are available from some sources. It also seems that these checks are useful when retrieving a whole document and not just a fragment. With the current proposal I could use (the somewhat obscure): http://example.com/text.txt#char=0,;length=9876 to apply an identity check to a whole document. It also seems that I might want to merely identify the required charset of the whole document, but I can not do so without specifying either a length or an md5 hash. Rather than just rework the phrase "hash sum" to reduce the "astonishment," I would hope that would be possible to make these two separate text/plain add-on features more independent and even allow for the extension of the "identity check" feature in the future to more than just length, md5 and charset. -- Bill McQuillan _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf