HP review for LC

Brian asked me to do the HP review of the editor's draft for the last 
call vote.

The version reviewed was: 2007/02/21 16:25:45 number 1.228

Brian asked me to limit my review to substantive issues, and not merely
editorial matters.

Overall comment is this is a good document; the normative text is
particular clear; and this is ready for last call.

I suggest that the following three changes should be made before last call.

1) IRI throughout

Globally substitute URI with IRI, and change normative ref
to RFC 3986 to RFC 3987.

It is highly likely that this suggestion will be made by last call
reviewers, so it is better to address the comment before LC.

2) s/transformation/"transformation"/

In the first normative box in section 6 (http headers)
replace [[transformation]] with [["transformation"]] i.e. add quotes.

Both values are legal according to the syntax of [HTTP-LINK], the WG intent
is, I believe, a quoted string, not a token.

3) clarify normative text

A silly pedantic point is that while the textual conventions for
normative text are very clear and clean, that are not followed

Also, since I initially reviewed a black/white printed version,
I missed the coloured background for Paras 1 and 3 of section 8.

It would be good to have all normative text clearly and uniformly
distinguished throughout by being in the boxes with coloured backgrounds.
The headings "Normative Statement" etc. should either be for
all such boxes or none.


In addition, I make the following observation about caching, and
one further comment, that maybe should result in changes before last call.

4) caching

Recent experiments trying to improve the security compliance of the Jena
GRDDL reader surprised me. During the application of a transform to
an XHTML document, the default behaviour of the XML parser I am using
is to contact w3.org to get the DTD (I think).
This suggests that the text in section 3, concerning namespace documents
and caching, is not sufficiently general. DTDs presumably get a lot of
hits if this is standard behaviour when parsing XHTML.
It also points to difficulties in implementing this informative text.
A higher level Web technology such as GRDDL, depends on lower levels.
All levels of the Web use URIs, and it may not be feasible for a GRDDL
implementation to have full control over the caching behaviour of all
its subsystems.

5) validation and xmlns on XHTML

While conducting these experiments,
I was surprised in that some tests started failing when I changed the XML
parser configuration to stop validation completely, (and hence not fetch 
the DTD).
The reason is that a transform such as




which explicitly requires that the elements html, head and title be in
the appropriate namespace. The namespace declaration is in the DTD.
If all DTD validation is switched off, then these elements are not
namespace qualified, and do not match the select, so that the output
gives an empty title.

This probably could be fixed by modifying the text in the penultimate
para of section 7 to require that XHTML documents are read as in the
appropriate namespace. This may require some change to the normative text.



Received on Tuesday, 27 February 2007 18:58:01 UTC