W3C home > Mailing lists > Public > uri@w3.org > August 2004

Re: Relative URI or relative URI reference

From: Elliotte Rusty Harold <elharo@metalab.unc.edu>
Date: Thu, 19 Aug 2004 07:24:13 -0400
Message-Id: <p06010201bd4a3c94ce6d@[]>
To: Roy T.Fielding <fielding@gbiv.com>
Cc: uri@w3.org

At 3:02 PM -0700 8/18/04, Roy T.Fielding wrote:
>>  Yes, there is. If it is confusing (as this blatantly is), then the 
>>confusingness should be noted in the spec. Not doing so makes the 
>>spec harder for the average person to understand.
>What confusion?  So far, the only thing that is confused is that
>some people believe use of the term relative URI cannot exist
>separately from the word "references".  Nobody seems to be confused
>about what a URI may be, nor are they confused about what a Relative URI
>may be, so the request to change all occurrences of "Relative URI" to
>"Relative URI Reference" and <relative-URI-reference> is both
>editorial in nature and fundamentally wrong.

You aren't listening to the same people I am then. Many people are 
very confused about what a relative URI is, and in most contexts 
outside the W3C (and some inside it) the claim that relative URIs are 
not in fact URIs would meet with puzzled stares. The common usage is 
that there are two kinds of URIs, relative and absolute (well, really 
the common usage is that there are two kinds of URLs, relative and 
absolute; and "URI" is just a funny spelling of "URL" you find in W3C 
specs). Honestly, I would prefer that this be the approach taken by 
2396bis and the whole notion of URI references be left in the dust, 
but that's been shot down in the past so I thought maybe we could at 
least be clearer about the distinction between a URI reference and a 

>That section is defining syntax.  <relative-URI> is a protocol element
>that is used by this specification and other Internet specifications
>as the means of syntactically distinguishing between a URI and a
>relative URI.  The text of the document is already quite explicit
>and I am not going to make this change because it would require
>massive reformatting of the ABNF rules and create an arbitrary
>disconnect between this specifications and all prior specifications
>of the standard.  That is not worth it even if I were to accept the
>premise that it is confusing, which I do not.

I see no problem with redefining the names of the ABNF productions in 
a new version of the specification. The relevant question is whether 
the same strings do or do not match the complete ABNF, not how the 
pieces of the ABNF are named. Of course, people do read the specs, 
and unlike machines they do impart unjustified weight to the names of 
the productions, so it's sensible to make sure the names do in fact 
correctly describe the set of strings identified by the production.

   Elliotte Rusty Harold
   Effective XML (Addison-Wesley, 2003)
Received on Thursday, 19 August 2004 13:34:45 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:08 UTC