Re: SVG12: <iri>

On Monday, April 18, 2005, 12:59:15 AM, Bjoern wrote:

BH> Dear Scalable Vector Graphics Working Group,

BH>   In the
BH> <iri> data type is not properly defined.

We find that it is well defined, with a normative reference to an IETF
Draft Standard.

BH> The draft refers to this type for both the CSS url() functional
BH> notation as well as literals.


BH> It is also unclear how SVG Tiny 1.2 is in a position to re-define
BH> normative dependencies like CSS 2.0 which does not allow anything
BH> but URIs in the url() notation,

Looking at the normative reference in CSS 2.0


    "Uniform Resource Identifiers (URI): Generic Syntax and Semantics",
    T. Berners-Lee, R. Fielding, L. Masinter, 18 November 1997.
    Available at
    This is a work in progress that is expected to update [RFC1738] and

we notice that there is indeed a problem - it has a normative reference
to a long-expired Internet Draft which,as you know, has a shelf life of
6 months.

The "expected to update" language, while woolly, leads us to believe
that a forward reference to the most current specification would be
intended. draft-fielding-uri-syntax-01.txt eventually became RFC 2396
which was obsoleted by RFC 3986 which is compatible with IRI, RFC 3987,
provided the escaping mechanism in section 3.1 of RFC 3987 is carried

Thus, while a strict conformance to whatever the long-expired ephemeral
draft said is difficult at this juncture, we believe that the intention
is fairly clear when used in an XML context - as with XLink, IRIs are
used in the XML content and may be escaped for retrieval if the protocol
does not support IRI directly.

For external stylesheets in the CSS syntax, ie not in an XML context, a
careful examination of the parsing of that format would be required to
determine if there is any incompatibility. However, SVG Tiny 1.2 does
not make use of external author stylesheets.

BH>  and other specifications like xml:base and XLink 1.0
BH> do not use IRIs either

Their definition predates the issuing of RFC 3987 but they were intended
to use the same syntax. Now that RFC 3987 has been issued,
specifications are being updated to remove the copy-paste versions of
the escaping mechanism and to refer to RFC 3987 directly. SVGT 1.2 does
this also.

BH>  (and IRIs are incompatible with their URI I18N mechanisms).

We find that this is not the case, and that considerable liaison effort
took place between the authors of RFC 3986 and RFC 3987 to ensure that
they were compatible.

BH> Please change the draft such that there are different data
BH> types for IRI literals and IRIs in url()

We decline to do so, for url() usage in an XML context, as this would
introduce needless inconsistency in IRI usage. For external stylesheets
in SVG 1.2 Full, we will consider your comment and see whether there is
a problem and if so, what to do about it.

BH> and precisely explain the impact of SVG's re-definition of its
BH> normative dependencies in this regard.

We find that introducing a normative dependency on an expired 1997
internet draft which no longer exists would be problematic, and trust
that you do not, in fact, suggest this. We plan to continue referencing
the relevant latest specifications on the IETF standards track instead.

Please let us know within two weeks if this does not address your

 Chris Lilley          
 Chair, W3C SVG Working Group
 W3C Graphics Activity Lead
 Co-Chair, W3C Hypertext CG

Received on Thursday, 13 October 2005 17:39:48 UTC