Re: Must clients use all transformations specified?

On Tue, 2006-09-05 at 12:20 -0400, Chimezie Ogbuji wrote:
> On Tue, 5 Sep 2006, Dan Connolly wrote:
> 
> >> However, I do think we should clarify that either the client should either:
> >>
> >> 1) If it can run a transforms, it runs the transform.
> >>
> >> 2) Or if can run a transform,  the client may or may not run the transform.
> >
> > I think the spec already says 2) clearly enough.
> 
> 1>
> > I think introducing conformance labels like "GRDDL client" is
> > much more trouble than it's worth.
> 
> Than is my action item regarding defining such labels a moot point?

I think it's good to have an explicit discussion of whether
to have conformance labels. I hope to convince you that "GRDDL client"
is not worth doing, but maybe you'll convince me (or a critical
mass of the WG) that it is.

(I'm not quite sure what "moot point" means; I have a vague
memory that, like "table", it has opposite meanings in the U.S.
and the U.K. I'd look it up, but I don't think it matters that much.)


> > I think specifying the meaning of documents and publishing
> > examples and test cases is sufficient (as well as necessary).
> >
> > Given the security issues around running code published
> > in the Web, which transformations to run clearly must
> > be left up to local policy, no?
> 
> I'm not sure.  I mean, I can see the security issues being a prime factor 
> in having the 'client' decide which transformations to run, but assuming 
> there were no issues with 'local' policy, I don't understand how you can 
> claim the transformations (identified by the publisher) 'preserve' the 
> author's meaning but leave it up to the client which transformations to 
> run.

Hmm... I thought the current draft was reasonably clear
that GRDDL is about extracting _part_ of the meaning of a document,
since, in general, documents have more than RDF data in them.
The intro tries to make the point...

[[
Data formats like XML and XHTML are used in the Web for a large spectrum
of purposes, from poetry and drama to spreadsheets and databases. The
information in a poem may be rich and subtle; we might use a computer
pick out the author's name, but themes and opposing forces are not
readily computable.
]]
 -- http://www.w3.org/2004/01/rdxh/spec#intro

and another examples says...

[[
It shows that the RDF data extracted via the dc-extract.xsl
transformation is part of the information contained in the document
]]


The namespace document includes a (some of) qualification...

[[
transformation
          
    relates a Document to an
              Algorithm,
              usually represented in XSLT, for
              extracting an RDF/XML
              representation of (some of) the document's
              meaning.
]]
 -- http://www.w3.org/2003/g/data-view#transformation


but you're right that the most relevant part of the spec does not...

[[
A reference to http://www.w3.org/2003/g/data-view from the profile
attribute (c.f. section 7.4.4.3 Meta data profiles of [HTML4]) of an
XHTML document[XHTML] indicates that links of type transformation relate
the document to transformations that preserve its meaning.
]]
 -- http://www.w3.org/2004/01/rdxh/spec#grddl-xhtml


I added a @@TODO there in 1.85.



> Perhaps, a clear usecase as to the value of choosing to only run a subset 
> (besides security considerations) would help me understand the value in 
> this.

I think security considerations are sufficient reason.

>   Looking at the xml-stylesheet processing instruction specification, 
> it doesn't seem to leave the matter up to the client.
> 
> [[[
> 
> Multiple xml-stylesheet processing instructions are also allowed with 
> exactly the same semantics as with LINK REL="stylesheet".
> 
> ]]]

Note that says nothing about conformance of software; it only
specifies meaning of documents.

> I'm not aware of what is expectred of the client from a 
> link/@rel='stylesheet' in an XHTML document (does it have the option to 
> not apply the stylesheet?).

Of course.

When you say "what is expected", are you referring to market
expectations or to conformance requirements? My sense of the
state of the market is that a browser that doesn't do stylesheets
unacceptable in typical usage. But there are atypical uses
(limited compute resources on a mobile device, maybe? text-only
browsers?) where not using stylesheets makes sense.

So the Right Thing happens even thought the spec doesn't
mandate that software do anything in particular.


> > Consider an agent that has a hard-coded list of profile
> > and/or transformation URIs; its policy is to execute
> > those transformations and no others.
> > For example, a big data aggregator (think: yahoo local, ...)
> > might publish a list of 20 profiles and transformations
> > (hCard, eRDF, ...) and offer to aggregate data that uses
> > those profiles. They might add new ones over time, after
> > carefully reviewing and caching the published XSLT implementation,
> > or by re-implementing the transformation in C locally.
> >
> > Is anybody interested to make a test case to illustrate
> > that case?
> 
> I might, mostly because I'm concerned that there is a conflict with 
> claiming the transformation preserve the document's meaning (which is a 
> vague statement in itself) and not mandating that all referred 
> transformations should be run (barring local such policies)

Er... again, no fair "barring local policies"; local policies
are my main point.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Tuesday, 5 September 2006 16:53:05 UTC