[XML-SIG] DTD for recipes

[Walter Underwood]

At 11:27 PM 3/11/99 -0500, A.M. Kuchling wrote:
>I've taken a first whack at a DTD for storing recipes.  It's nowhere
>near complete, and there are a bunch of open issues.  Please see
>
>	http://starship.python.net/crew/amk/recipe/
>
>for the DTD and a little sample document.  As we did with XBEL, we can
>discuss the DTD here on the XML-SIG list.

Good start. Here's a search engine perspective on metadata: 

   <!ELEMENT recipe (title, author?, source?, [...]

I recommend using the Dublin Core elements as needed, with the
formats of the content specified. These three are particularly
important for search engines, since they are the ones that are
shown in the results pages:

   title
   description
   date    (use the w3.org subset of ISO 8601, that is, "1999-04-12")

Secondary:

   subject (similar to the keywords metatag, but might use an 
            organized vocabulary, like Library of Congress Subject
            Headings)
   creator (instead of author, above)
   contributor (for editor, illustrator, etc.)
   publisher
   language (use the HTTP/ISO codes, like "en-GB" for British English)

and so on. For personal names (creator, contributor), I'd recommend
choosing one of two methods:

   signature form               AACR2 (library cataloging std)
   --------------               ------------------------------
   Walter R. Underwood          Underwood, Walter R.
   Margittai Neuman János       Von Neumann, John
   Neumann János                Von Neumann, John
   John von Neumann             Von Neumann, John
   Tanizaki Junichiro           Tanizaki, Junichiro
   [unicode for above]          Tanizaki, Junichiro
   Clemens, Samuel              Twain, Mark

And as this example shows, don't make elements for "firstname"
and "lastname", but you might want to allow for multiple creator
elements, with and without romanization, as well as for 
multiple authors.

Finally, I think that Dublin Core is perfectly adequate for
this application. RDF is too complex (or perhaps inadequately 
explained) for the average author or developer, and the HTML 
conventions (title, keywords, description) are just not enough 
information. Dublin Core is  the "Goldilocks" solution, just right.

wunder




--
Walter R. Underwood
wunder@infoseek.com
wunder@best.com (home)
http://software.infoseek.com/cce/ (my product)
http://www.best.com/~wunder/
1-408-543-6946