Pete writes:
> Tom,
>
> Can we make these changes as an errata note in the current (2008/08/04) document please?
>
> Or would you prefer we create a new version (2010/03/nn)?
>
> The former would mean no change in "instance" XHTML documents (and hopefully the transform will just start working again for those documents); the latter would require people to change their XHTML documents to cite the new profile URI and would leave things "broken" for documents citing the current profile URI (at least for cases where the GRDDL processor doesn't grab the DTD).
>
> My preference would be to do the former.
I agree with your preference. It is the sort of change for
which we have used Errata notes in the past -- a practice about
which I have heard no complaints. It fixes the problem with
the least disruption.
> (iii) a note (another addendum? A new appendix?) should be
> added to the profile document to make it clear that "instance"
> documents referencing the DC-HTML profile should avoid the use
> of character entity references (other than the five predefined
> by XML (&, <, >, " and ')).
Rather than bury the point in an appendix, I would suggest
adding a sentence of explanation about the issue near the top
of the document, e.g., at the end of the introduction. This
sentence could be clearly marked and dated as an addition and
could a link to the Errata section, which could marked with an
anchor following [2] - e.g., "#errata" - with a full description
of changes made.
Does anyone out there see good reasons why a new version should
be created to replace [1]?
Tom
[1] http://dublincore.org/documents/2008/08/04/dc-html/
[2] http://dublincore.org/documents/2008/08/04/dc-html/#ack
On Thu, Mar 11, 2010 at 12:16:06PM +0000, Pete Johnston wrote:
> > This suggests to me that, because a GRDDL processor may employ a non-
> > validating XML parser, it is inappropriate ("inherently unsafe") to use
> > character entity references like and © in a document to be
> > processed (in this case both the DC-HTML profile document itself and
> > "instance" documents), and so numeric character references (which a
> > parser, validating or non-validating) can interpret without reference to
> > an external source like a DTD) should be used instead.
> >
> > That's the conclusion I'm coming to, anyway. But I've stared at this so
> > long that I may be completely missing something else which is the real
> > source of the problem. Can anyone confirm/deny or otherwise help out with
> > this please?
>
> I've had it confirmed to me off-list that this is indeed the source of the problem.
>
> So I think a few things need doing:
>
> (i) the profile document needs to be altered to use numeric character references instead of character entity references
> (ii) any fragments it references via SSIs need to be altered to use numeric character references instead of character entity references
> (iii) a note (another addendum? A new appendix?) should be added to the profile document to make it clear that "instance" documents referencing the DC-HTML profile should avoid the use of character entity references (other than the five predefined by XML (&, <, >, " and ')).
>
> http://www.w3.org/TR/2008/REC-xml-20081126/#sec-predefined-ent
>
> I don't think any change is required to the XSLT transform itself.
>
--
Thomas Baker <[log in to unmask]>
|