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

Re: Relative URI or relative URI reference

From: Roy T. Fielding <fielding@gbiv.com>
Date: Thu, 19 Aug 2004 15:02:35 -0700
Cc: uri@w3.org
To: Sam Ruby <rubys@intertwingly.net>
Message-Id: <7A0925F4-F22B-11D8-BD91-000393753936@gbiv.com>

On Thursday, August 19, 2004, at 01:58  PM, Sam Ruby wrote:
> I believe that this is an example of what Paul was referring to:
>
>   http://www.imc.org/atom-syntax/mail-archive/msg08608.html
>
> To summarize: the fact that relative URIs references are not 
> considered URIs is a new position, and should somehow be highlighted.

Umm, wrong.  That person apparently hasn't read RFC 2396 or rfc2396bis,
since neither one supports his opinion.  RFC 2396 did not define a
syntax rule for "URI" -- only absoluteURI and relativeURI.

    Relative URI references are distinguished from absolute URI in that
    they do not begin with a scheme name.  Instead, the scheme is
    inherited from the base URI, as described in Section 5.2.

One of the very first issues that came up with regards to the revision
is whether we should clarify this topic in rfc2396bis, which is why URI
was added to the specification.  As I said, it has already been 
clarified
and this discussion is just a hangover.

>> If you want to speak for yourself -- do you personally think the
>> wording in RFC2396bis is confusing? -- then that would be
>> fine, too.
>
> I personally find the following confusing in RFC296bis:
>
>  - - -
>
> The title of section 4.2:
>
>   Relative URI
>
> I understand and appreciate the argument that the BNF needs to retain 
> the previous usage of the term (some informative text making this 
> history clear wouldn't hurt), but the title shouldn't reinforce the 
> now incorrect usage of this term.

There is nothing incorrect about that term!  You are imagining some
mystical quality that simply does not exist.  "Relative URI" does
not imply anything about the definition of URI.  We call it that because
that is the term people look for when they say: "gee, I wonder what part
of this document talks about relative URI?"  That is why the section is
titled Relative URI, the following section is titled Absolute URI,
and the parent section is titled Usage.  Those things are still called
"relative URI" in common practice, technology, specification, and
implementation.  They *should* be called that.  A "relative reference"
is an instance of the relative-URI syntax *when* it occurs inside a
URI-reference protocol element.

The text of those sections, and the associated ABNF, defines
those protocol elements in a way that is unambiguous.  Changing the
titles and the ABNF element just makes it harder for implementers
to find the definitions.

> The fifth bullet of section 6.3:
>
>    Prevent dot-segments appearing in non-relative URI paths.
>
> I'll suggest dropping the words "non-relative" from this text.

Yes, that is a bug.  A suggestion was made at the IETF meeting to
drop the entire section on the basis that it just repeats most of
what has already been said better elsewhere.

....Roy
Received on Thursday, 19 August 2004 22:02:20 UTC

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