- From: Adam M. Costello BOGUS address, see signature <BOGUS@BOGUS.nicemice.net>
- Date: Sat, 13 Mar 2004 06:37:01 +0000
- To: uri@w3.org
Tony Hammond <tonyhammond@mac.com> wrote: > > Yes, it would be better to use either the generic authority syntax > > or the URN authority syntax for new URI schemes that make use of > > delegated naming authorities. > > There are problems with both of these approaches which have been > carefully documented in the FAQ (available at <http://info-uri.info/> > and presented earlier on this list when the -00 draft was posted. I have now read the FAQ. I looked for a description of problems with the generic-authority approach, and this was the closest I could find: Location dependence - URI schemes are variously categorized as location-dependent or location-independent with respect to whether the URI contains a network authority component. The FAQ consistently puts the word "network" in front of the phrase "authority component", as if the //authority component of a URI can refer only to network entities. But the authority component is not restricted to network entities, and was introduced in RFC-2396 to remove that restriction. RFC-1738: //login (IP-based user@host:port) RFC-1808: //net_loc (same thing, renamed) RFC-2396: //authority authority = server | reg_name (server is another renaming of net_loc/login, reg_name is new) It seems to me that reg_name was introduced specifically to hold names like info's namespace component. (I just wish it had been introduced in a way that avoided type-ambiguity. In foo://bar/, is bar a hostname or not? You can't know unless you know the foo scheme. I wonder if it's not too late to fix that. Even while I'm arguing that info: ignores the intent of RFC-2396, I worry that conforming to that intent would create problems.) "Ray Denenberg, Library of Congress" <rden@loc.gov> wrote: > Please look at: http://info-uri.info/registry and consider these two > entries: 'lccn', and 'srw'. Lccn has no authority component because > the info:lccn just takes an existing lccn, normalizes it, and sticks > it in a uri. No authority, because nobody is coining the URI. I don't see how that matters. RFC-2396 says: Many URI schemes include a top hierarchical element for a naming authority, such that the namespace defined by the remainder of the URI is governed by that authority. I think that describes lccn as well as it describes srw. The string "lccn" is mapped by the info registry to something (a spec or an organization or a series of specs issued by an organization or whatever) that governs the syntax and semantics of the rest of the path. By the way, there is something screwy about the web server at info-uri.info. Apparently, if the User-Agent request field is absent or begins with anything other than "Mozilla", a cryptic XML error page is returned. AMC http://www.nicemice.net/amc/
Received on Saturday, 13 March 2004 01:37:02 UTC