Re: Should our implementation be updated to conform?

David (and Werner and Eric),

I will take your comments/questions back to the XSL FO subgroup
for further deliberation.

Unofficially (and without consultation with others), I'm pretty sure 
it's not an unrecoverable error to have an unquoted <uri-specification>.
At the very least, an unquoted <uri-specification> falls under
the case of the Note added to section 5.11 by the Errata [1]
that says that, by way of recovery, an implementation may treat
the unevaluated property value as a string.  So I wouldn't change
your implementation to make this an error.

paul

[1] http://www.w3.org/2001/10/REC-XSL-20011015-errata


At 10:28 2003 02 23 +0400, David Tolpin wrote:

>> >
>> >Please note that it does not accept a <string>, but an <uri-specification>
>> 
>> Per http://www.w3.org/TR/xsl/slice5.html#section-N8794-Property-Datatypes
>> a <uri-specification> is "A sequence of characters that is "url(", followed..."
>> and <string> is "a sequence of characters".  Therefore, the spec is
>> saying that <uri-specification> is a subset of the <string> datatype.
>> 
>> I agree that the spec could be clearer in the area of property datatypes,
>> but I hope my explanation clarifies how our response to your comment doesn't
>> contradict the spec.
>
>Paul,
>
>the Recommendation says that a string may be represented either by a literal,
>in which case it must be taken into quotes, or by an enumerated token. An enumeration
>token is an NCName, and qoutes are not used.
>
>Since uri('http://www.w3.org/TR/xsl') cannot syntactically be a NCName,
>should our implementation, RenderX XSL Formatter XEP be updated to require
>that every uri specification is taken into single or double quotes (around
>the whole thing) with optional quotes inside?
>
>It currently erroneously allows to omit outer quotes around uri-specification
>and simple write
>
>external-destination="uri('http://www.w3.org/TR/xsl')"
>
>Now I realize that since uri(...) is a special case of a string per its definition
>(since both are defined to be kinds sequences of characters, then the latter is
>a special case of the former), then this approach is wrong.
>
>The right approach would be to write
>
>="'uri(http://www.w3.org/TR/xsl)'"
>
>or
>
>="'uri(&quot;http://www.w3.org/TR/xsl&quot;)'"
>
>And the case without outer quotes is an error because if b and c are subsets of a
>then c is subset of b.
>
>Thank you for clarification,
>David Tolpin
>RenderX