RE: DSig comments on XML Base

At 00/07/27 13:18 -0700, Jonathan Marsh wrote:
>John wrote:

> > Moreover, the 'Included if validating' terminology does in fact reference
>by
> > hypertext link the definition of 'included', so my use of that section's
> > phrasing is indeed appropriate, and it again brings up the question:
>
> > Should we be introducing an attribute beginning with xml: that seems to
> > violate section 4.4.2 of the XML 1.0 spec?
>
>Personally, I think trying to use the XML spec to justify this issue one way
>or the other is a waste of time, since the XML spec does not specifically
>address this issue and extrapolating existing text does not necessarily
>reveal the authors' original intentions (if they even had any).
>
>Previous comments on this subject (led by Martin Duerst) have shown that
>both alternatives are coherent, but that neither is optimal for all cases.
>XSLT sets up a historical presumption that elements and processing
>instructions in external entities should retain an independant base URI.

Sorry, but I wasn't aware of the text in XML that John has brought up.
While I don't think it's always easy to second-guess what spec writers
had in mind, and I therefore agree that such second-guessing should
be limited as much as possible, I don't think at all that the section
about inclusion in 4.4.2 needs interpretation. It quite clearly says
that including an (external) entity means doing a textual replacement.
To read it otherwise would be very surprising, at least to me.

XML therefore sets up a historical presumption that there is no
difference for an XML processor between an external entity and
an XML document where this entity is inserted.
It seems therefore that the question is not whether XBase follows
XSLT or not, the question is how to make sure that XBase follows
XML, and how we can make sure that other specs do the same.


>C14N could insert xml:base attributes on elements coming from the top-level
>of an external entity to fix the problem in most cases.  The only case not
>coverred is where an external entity has a processing instruction at the top
>level.

You mean that there could be a relative URI in the PI?
Interesting case. But I think a PI has to be treated in
the same way as free text. Then either the base of the
external entity also applies to free text, which would
be in disagreement with XSLT, and in which case we have a
much more serious problem, or the base of the external
entity doesn't apply for both cases.


Regards,   Martin.

Received on Friday, 28 July 2000 18:56:22 UTC