Re: Comments on GRDDL (using 3rd-party XML schemas with GRDDL) [OK?]

On Mon, 25 Jun 2007 14:21:32 -0500, Dan Connolly wrote:

 > On Mon, 2007-06-25 at 13:48 -0400, Andrew Eisenberg wrote:

 >> Hi Dan,

 >> The WGs believes that the specification supports both the use
 >> of schema documents that are retrievable from the namespace
 >> name and schema documents that are pointed to from the
 >> document itself.

 > Right.

 >>  We do not believe that there a provision for using other
 >>  schema documents (e.g. schema documents retrieved from a
 >>  local repository).

 > Right.

Thanks for the clarification.

 >>  If this capability is not provided, then we believe that this
 >>  should be mentioned in the specification.

 > In my 30 April reply,
 > http://lists.w3.org/Archives/Public/public-grddl-comments/ 
2007AprJun/0038.html
 > I explained that 3rd-party claims are not among the use cases
 > GRDDL is designed for. And yet you ask again that we mention
 > it, and again without any justification. There is an infinite
 > number of use cases that GRDDL is not designed to cover; I
 > think that's implicitly clear. I don't see any reason to
 > mention this one in particular.

Sorry for having been unclear.  We assumed that the reason would
be obvious: the failure to support schema documents from
arbitrary locations will be confusing to some readers who, like
us, come to the GRDDL spec having some familiarity with the XML
Schema spec and its implementations, and we believe that you
should avoid confusing your readers without need.

Schema documents used for validation or annotation are not always
mentioned explicitly in the document to be validated and cannot
necessarily be found by dereferencing the namespace name (since
the namespaces spec offers no guarantee that it can be
dereferenced).  Even when some schema documents CAN be found that
way, those are not necessarily the schema documents the user
wishes or needs to use.

Users of XML Schema accustomed to using schema documents to guide
the annotation of document instances are likely to be surprised
by the failure of a spec like GRDDL to support a common use case.
The assumption that GRDDL annotations will only be available, or
only of interest, when present in schema documents at the
namespace name, or named in the document itself, is a surprising
and unnecessary form of closed-world assumption.

Note also that the location of the schema document has nothing to
do (as far as we can tell) with whether the annotations in
question stem from first, second, or third parties.  If
annotations are found in a schema document found at the namespace
name, it's probably safe to infer they stem from the owner of the
namespace. It is clearly unsafe, however, to assume the converse:
the presence of annotations in a schema document found elsewhere
does NOT license the inference that those annotations do not stem
from the owner of the namespace -- or, for that matter, from the
creator or owner of the document being annotated.

To be more blunt, the omission from GRDDL of support for GRDDL
annotations in arbitrary schema documents is a design flaw.  It's
fixable, and we think you should fix it.  If you choose not to
fix it, you can improve your spec by making clearer that the
design you present chooses, for reasons you may or may not choose
to elaborate, not to support use cases which some of your readers
will expect you to support.

 > If you provided some justification for your request, I could
 > perhaps consider it further with the Working Group.

Our apologies for not being more explicit in our earlier comment.
I hope this not makes the justification for the change clearer:
it will make your spec less confusing to readers you should care
about.

 > As it is, I hope you find this response satisfactory.

Unfortunately not.

--C. M. Sperberg-McQueen
   for the W3C XSL and XML Query Working Groups

Received on Wednesday, 18 July 2007 00:52:01 UTC