[OAI-implementers] Branding

[Simeon Warner]

My initial reaction to the question of what a style-sheet specified in
a metadataRendering declaration should do is: produce fragment XHTML.

Fragments:

As Hussein said, fragments are appealing from the point of view of
integration. However, if there really are two useful cases, fragment and
standalone, we could provide some way to declare this outside of the mime
type (c.f. standalone = "yes" | "no" in XSLT output declarations:
http://www.w3.org/TR/xslt#output). It seems hard to tell given almost no
use of this feature so far.

While I understand Jeff's concern about running XSLT processes on one's
server, I don't see any real difference between a service provider doing
that (perhaps in a well-controlled sandbox) or asking a downstream user to
do it in their browser. One might even argue that service provider is
likely better equipped to control the dangers than a downstream user?

The branding spec says that "the style sheet must be able to render the
metadata part of a GetRecord response" which is rather vague (for which
I'm partly to blame). Should this say instead that "the style sheet must
be able to render the <metadata> part of a <record> response" meaning that
one could give it a record from a GetRecord or a record from a ListRecords
response? The latter sounds more useful to me.

XHTML:

This seems the obvious default at present. However, XSLT1.0 doesn't have
an XHTML output method (XSLT2.0 does but I hear tools for that aren't
ready for prime-time yet, and the spec is just a Working Draft). With XSLT
1.0 I think that means one should probably use the 'xml' output method to
avoid problems with possible collapsing of empty elements to ill-formed
xml (e.g. <p/> -> <p>, see:
http://www.w3.org/TR/xslt#section-HTML-Output-Method)

I agree with Hussein that it would be sensible to add an outputMimeType
attribute to the <metadataRendering> element. Can anyone see a reason not
to do this? (I should also fix the mimeType type in the schema which
has a pattern that is too restrictive to permit all mime types).


Incidentally, I added a metadataRendering declaration to
http://arXiv.org/oai2 so I suppose there are now 5 repositories that do
it.

Cheers,
Simeon




On Sat, 2 Oct 2004, Jeffrey A. Young wrote:
> I vote against fragments. The reason is a little involved.
>
> The problem starts with the fact that I am not willing to run unknown
> stylesheets on my machine. As mentioned at the bottom of section 5.4 of the
> XSLT spec (http://www.w3.org/TR/xslt#section-Applying-Template-Rules), there
> is a denial of service risk associated with running stylesheets.
>
> The solution I've come up with is to create an XSLT processor web service
> coupled with a catalog of stylesheets that the processor service is willing
> to perform. This has been bundled into an open-source Java Servlet
> distribution named XSLTProc so anyone can set it up to support stylesheets
> they are interested in using. The parameters to this service are the same as
> they would be calling the XSLT processor from the command line.
>
> The catalog for this service will be OAI-harvestable which will allow me to
> create a union catalog and couple that with another open-source web service
> that accepts the same arguments XSLTProc does, except that instead of
> performing the transform, it will perform an HTTP redirect to a willing
> XSLTProc service instead.
>
> This solution may be less efficient than running the stylesheet directly,
> but it is the only way I can think of to manage the risk of running unknown
> stylesheets. There are many other interesting and uses for this model,
> though, that I won't go into now.
>
> Getting back to the problem with fragments, a stylesheet that produces
> fragments would be fine if I imported it locally into another stylesheet,
> for example, but I think the XSLTProc service must be bound to produce
> well-formed XML so that the client can manage the response using XML tools
> rather than being forced to treat it as a string.
>
> Sorry that this has led to such a tangent, but this is the problem and
> solution I am facing.
>
> Jeff
>
> > -----Original Message-----
> > From: oai-implementers-bounces at openarchives.org [mailto:oai-implementers-
> > bounces at openarchives.org] On Behalf Of Hussein Suleman
> > Sent: Saturday, October 02, 2004 12:48 PM
> > To: Young,Jeff
> > Cc: oai-implementers at oaisrv.nsdl.cornell.edu
> > Subject: Re: [OAI-implementers] Branding
> >
> > hi Jeff
> >
> > i would vote in favour of standalone XHTML 1.0.
> >
> > at first fragments sounded nicer (for integration), but then what if the
> > rendering is PDF or Mobile-HTML or some such thing ... and then again
> > XHTML is sooner or later going to be v2.0 anyway. with a complete
> > document, we can at least make the assumption of a mime type and promote
> > this to a problem the browser mediates.
> >
> > then maybe in a later "bugfix", Simeon (or whoever is the maintainer)
> > can introduce a "target-mime-type" attribute or some such mechanism in
> > the metadataRendering container.
> >
> > but is there a mime type for HTML fragments? if there is, then maybe the
> > fragment is the way to go ... as long as down the road we can widen the
> > scope without any changes. if there is no fragment mime type, maybe we
> > can introduce one :)
> >
> > ttfn,
> > ----hussein
> >
> >
> >
> > Young,Jeff wrote:
> > > I have an immediate need for a best-practices description for creating
> > XSLT
> > > stylesheets for use with the description/branding/metadataRendering
> > element
> > > in an OAI Identify response
> > > (http://www.openarchives.org/OAI/2.0/guidelines-branding.htm).
> > >
> > > The only guidance this specification gives regarding the stylesheet
> > itself
> > > is "the style sheet must be able to render the metadata part of a
> > GetRecord
> > > response for the specific metadata format". Unfortunately, it doesn't
> > say
> > > what the stylesheet should render it into. Should it be an HTML fragment
> > > suitable for dropping into the body of an HTML document? Should it be a
> > > stand-alone XHTML document ready to display in a browser? Can I depend
> > on
> > > the result being assigned to a particular namespace? There are limits to
> > how
> > > many variants I should have to attempt in order to parse the result and
> > use
> > > it.
> > >
> > > There are only 4 repositories registered at UIUC that currently use this
> > the
> > > metadataRendering element, two of which are mine
> > >
> > (http://gita.grainger.uiuc.edu/registry/ListReposWithDesc.asp?desc=brandin
> > g)
> > > . I am on the verge, however, of creating a system that depends on them,
> > as
> > > well as updating many of my repositories to include them.
> > >
> > > It would be nice to have this group's consensus before I commit to a
> > > particular solution.
> > >
> > > Jeff
> > >
> > > _______________________________________________
> > > OAI-implementers mailing list
> > > List information, archives, preferences and to unsubscribe:
> > > http://openarchives.org/mailman/listinfo/oai-implementers
> > >
> >
> > --
> > =====================================================================
> > hussein suleman ~ hussein at cs.uct.ac.za ~ http://www.husseinsspace.com
> > =====================================================================
> >
> >
> > _______________________________________________
> > OAI-implementers mailing list
> > List information, archives, preferences and to unsubscribe:
> > http://openarchives.org/mailman/listinfo/oai-implementers
>
>
>
> _______________________________________________
> OAI-implementers mailing list
> List information, archives, preferences and to unsubscribe:
> http://openarchives.org/mailman/listinfo/oai-implementers
>