RE: prs-/vnd- not broad enough; how about "ext-"

Hi,

I like best Sandro's suggestion in the last paragraph below
(any scheme that contains a hyphen-dash is non-IETF).  Mainly
because that reflects the current usage for 'trees' specified 
in RFC 2717 "Registration Procedures for URL Scheme Names".

From Section 3.2 "The IETF Tree" of RFC 2717, an excerpt:

  "The NAMES of schemes registered in the IETF tree MUST NOT contain the
   dash (also known as the hyphen and minus sign) character ('-')
   USASCII value 2Dh.  Use of this character can cause confusion with
   schemes registered in alternative trees (see section 3.3)."

From Section 3.3 "Alternative Trees" of RFC 2717, an excerpt:

  "The syntax for alternative trees shall be as follows: each tree will
   be identified by a unique prefix, which must be established in the
   same fashion as a URL scheme name in the IETF tree, except that the
   prefix must be defined by a Standards Track document.  Scheme names
   in the new tree are then constructed by prepending the prefix to an
   identifier unique to each scheme in that tree, as prescribed by that
   tree's identifying document:

      <prefix>'-'<tree-specific identifier>

   For instance, the "foo" tree would allow creation of scheme names of
   the form: "foo-blahblah:" and "foo-bar:", where the tree prescribes
   an arbitrary USASCII string following the tree's unique prefix."

Why do URL schemes need 'vnd' or 'prs' at all?

Cheers,
- Ira McDonald
  High North Inc


-----Original Message-----
From: Sandro Hawke [mailto:sandro@w3.org]
Sent: Wednesday, September 10, 2003 3:03 PM
To: uri@w3.org
Cc: draft-king-vnd-urlscheme-03.txt@roke.hawke.org
Subject: prs-/vnd- not broad enough; how about "ext-"




Re: http://www.ietf.org/internet-drafts/draft-king-vnd-urlscheme-03.txt

The text clarifies that the term "vendor" isn't actually intended to
mean "vendor".  

     > The term "vendor" is used in this document for simplicity.

It seems to me that this moves the complexity from the the RFC out
into the world of everyone using the schemes.  The term "vendor" would
be quite confusing (or even offensive) for organizations which clearly
set themselves aside from "vendors", such as consortia (eg W3C),
governments, universities, and industrial users.

My first idea is to mirror the DNS approach.  It's imperfect as well,
but at least people are used to it.  The W3C would use org-w3-*, MIT
would use edu-mit-*, the University of Manchester would use
uk-ac-man-*.  Of course, in this case, going back to "." instead of
"-" would be very nice.  And it all runs afoul of the fact that domain
names can be taken from people and organizations without their
consent.   (Java package names use this approach, more or less.)

An simpler approach is to just add the term "org" along side "vnd" and
"prs".

Even simpler: just one prefix that has no implications about what the
named-thing in the second place is.  Assuming that W3C wanted to make
a "web" URI scheme (just a joke, of course):
        open-w3c-web:foo
        free-w3c-web:foo
        a-w3c-web:foo
	by-w3c-web:foo
	tag-w3c-web:foo
	ext-w3c-web:foo    <<<===  My Favorite ("External/Extension URIs")
	user-w3c-web:foo

Finally, one could drop the prefix entirely, saying any scheme with a 
dash in it is a non-IETF one, and the first segment should be the
IETF-approved name of an organization controlling the rest of it.

      -- sandro

Received on Wednesday, 10 September 2003 18:56:11 UTC