- From: <noah_mendelsohn@us.ibm.com>
- Date: Fri, 24 Feb 2006 21:28:02 -0500
- To: Mark Nottingham <mnot@yahoo-inc.com>
- Cc: "Stuart Williams" <skw@hp.com>, www-tag@w3.org
Mark Nottingham writes: > [nit] I'd rephrase this slightly to reflect that URIs *always* encode > metadata for private purposes (e.g., where to find the file / what > code to dispatch to). I think that's a bit strong. Consider a uuid: scheme URI. I would say that it encodes metadata for private purpose only in the sense that you can tell the scheme name, and you can do literal equality testing to see whether two strings are the same URI. Except in that rather sterile sense, I claim it encodes no "metadata for private purposes" (You could in many cases crack the UUID to get a MAC address, but for practical purposes they are treated as opaque.) So, I think "always" is too strong, and interestingly so. > I think that it's just as important to make this point (i.e., on > equal standing with those you put forth): "Authorities MAY > communicate information about the structure of their URIs (in other > words, instructions of how to extract metadata from them) for use by > observers." > > Doing so is very useful and common, and people reading the opacity > documentation often conflate it with unhinted inference (which *is* > bad). I agree, though with the caveat that any consuming software the depends on knowledge of such structure is likely to be less general than software that either treats URIs as opaque, or that extracts only metadata covered by very widely deployed specs (e.g. extracting the scheme or hiearchy based on RFC 3986). Noah -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Saturday, 25 February 2006 02:28:10 UTC