Beefed up the demo a bit

[Alan Ruttenberg]

I've hooked in the mechanism by which the page is served as RDF with a
stylesheet so that html is displayed in the browser. I've only defined
this for pubmed at the moment and the specifics of the RDF will be
redone. However it demonstrates the basics. I've added, in order to
continue a discussion with Mark Wilkinson about how to reference the
LSRN registry, one way of doing this in the RDF, and a pointer to the
LSRN registry is shown in the html. I've also experimented with LSRN's
approach to embedding the web page for the record on the about page.

All this is just prototype, of course, for the purposes of discussion.
To have a look, go to http://demo.sharedname.org/n/r/pmid/16899496

-Alan

ps. the /r/ in the URL, instead of /record or /about is due to a purl
server glitch - I'll fix it when we upgrade that software next.

[Peter Ansell]

Hi Alan,

I thought the RDF/XML would be transformed by the XSLT, but really the
xsl stylesheet contains all of the information preformatted to HTML.
Certainly a novel way. Will this cause two requests for information
for each RDF file that is accessed, along with the 303 redirect that
would be 3+ requests for each resolution?

Generic transformation is hard even in the cases where you know what
predicates are given in the RDF/XML so it certainly eases that burden.

Cheers,

Peter

2009/5/15 Alan Ruttenberg <alanrut...@gmail.com>:

[Jonathan Rees]

Peter,

I agree with you that a generic transformation would be superior (more
scalable, more maintainable, more efficient, more transparent) to the
mechanism Alan has prototyped, no matter how attractive the results.
And I also agree that the generic transformation is harder than having
a specific one per databank.

Since you have experience with this, can you provide a few examples of
the presentation difficulties you've encountered with the generic
approach?

I don't think too much consideration should be given to efficiency, as
the HTML only matters in the unlikely even a person singly browses to
one of these pages. Large volume applications will only care about the
RDF, and will (we hope) be using an API or local mirror in any case.
But the other concerns still apply.

Jonathan

[Jonathan Rees]

I changed "Neurocommons" to "Neurocommons Project"

[Jonathan Rees]

Sorry all, my previous little message was meant only for Alan, who is
the only one who is expected to know what it is about. Please ignore
it. Once again I got thrown off by Google's annoying and
protocol-violating habit of setting Reply-to: to be the whole group
instead of the sender.

Jonathan

[Alan Ruttenberg]

On Fri, May 15, 2009 at 3:41 AM, Peter Ansell <ansell...@gmail.com> wrote:
>

> Hi Alan,
>
> I thought the RDF/XML would be transformed by the XSLT, but really the
> xsl stylesheet contains all of the information preformatted to HTML.
> Certainly a novel way. Will this cause two requests for information
> for each RDF file that is accessed, along with the 303 redirect that
> would be 3+ requests for each resolution?
>
> Generic transformation is hard even in the cases where you know what
> predicates are given in the RDF/XML so it certainly eases that burden.

As Jonathan points out, performance isn't really a concern. As to the
particular solution of using the wiki for the template, I'm more
concerned with the behavior and experience of using the shared names
at the moment, and with the process of coming to consensus on that.
Using the wiki template has the virtue that it's easy to change and
transparent as to what is being done viz the ease with which Jonathan
was able to fix something he didn't like. Once we get to the point
where we like the behavior we can concern ourselves with optimizing
for speed and scalability.

However, I would point out a couple of things to consider:
- Using this method for maintaining some databanks isn't incompatible
with using a more generic script for others,
- The wiki approach can be generalized so that there is a bit of
default inheritance so that databanks can use a generic page but a
more specific page can augment or override the default page
- The RDF doesn't need to literally be embedded in the page as I have
done - it could be generated from templates, or semantic wiki forms,
or any of a variety of mechanisms, from a simpler specification.
- It may be possible to write a little compiler that reads the wiki
page and generates an xslt from it, combining the best of both worlds.
We happen to have an expert in such matters on the team ;-)

Best,
Alan