Re: 2.6 patterns: let's try variations on the XML syntax

I heard a number of people at the metastructures conference last week
and at Jon's excellent XML Developer's Conference that followed it echo
the thought that the XSL query language that (indirectly) links
elements in an instance with styles or behaviour should be using
something compatible with RDF and XPointer and XLink and namespaces.

Or Perhaps I only thought I heard it because I believe it
so strongly myself.

The XML:XQL submission from AT&T may provide an interesting
alternative way of expressing queries.  If it goes anywhere, and if
the obvious incompatibilities with XML are removed [1], I can imagine
us implementing it in our products and using it for both queries and
for styles.  CSS is too hard for most people.  XSL has at the very
least to have a clearly defined subset that is very simple, much much
simpler than the current draft -- maybe a separate spec that's a
subset and thus less utterly overwhelming.  But as it stands, I can
easily see our customers coping with and even liking it.

[if this is too long a message, skip to the line marked XXX near the end]

In this spirit, Scott Lawton's idea of
    <xsl:match>
         <para></para>
    </xsl:match>
is interesting, although I think that, like most query-by-example
languages, it may not scale easily or elegantly to more complex
situations --
    For each SubSection S:
    Let e be the first block-formatted element directly contained within S;
    if the type of e is not Title
    then
	if S has an attribute "use-content-of-as-title",
	    and there is an element e2
		whose id attribute has the same value as
		<S>.use-content-of-as-title,
	then
	    insert <Title> before e, with the contents of e2
	else if S has an attribute "title"
	then
	    insert <Title> before e, with the contents of <S>.title
	else
	    insert <Title> with the contents "[no title given]"
	end if
    end if
might be hard to express that way, for example.  Is this style or
transformation?  As I have written it, it looks like a transformation,
but if you change
    insert <Title> with the contents "[no title given]" to
    make-paragraph size=18 weight=bold content="[no title given]"
then it's style -- I'm inserting <Title> so I can reuse the style of
the Title element, so I could also say
    apply the style for a <Title> before e to the content "...",
I suppose.

This is the sort of thing SGML people have been doing for years, and you
can do it declaratively (vide DSSSL) or procedurally (vide monde!),
and it makes no difference.



** XXX ** XXX ** The difficult challenge is this, then:


Make a style language, a link language, a locator language,
a transformation language and a resource description language [2]
that all feel as if they belong together, that are simple enough that
they can be used by relatively non-technical people in an afternoon,
and yet that are powerful enough to meet very advanced needs.


It may be that XSL is there, but I don't think so.  It's a draft,
and I think the wind blowing back is saying, make it simpler.  Make
it the same as other parts of XML.  Make it more accessible.

A lot of work has gone into it, and no-one said specs should be easy to
read by the intended user base of software conforming to a spec!
Many car drivers have little understanding of the chemical reactions
used to create motor oil, and nor should they have to.

The XSL query syntax needs to be integrated with XPointer and XLink,
or maybe a revised form of these, and XSchema, RDF and namespaces should
be built on that infrastructure.

I hope this doesn't sound too negative -- there is a lot of good work
in all of these specifications, but they really _do need to work together,
and to be seen to work together.

Lee



Notes:

[1] Using "." as an operator in a way that breaks if "." occurs in a
    name, using </> as if it were a legal construct; I've only had a
    chance to read it very quickly, over someone's shoulder, and will
    send comments back to the authors separately; it's very interesting.

[2] Actually I think RDF and XLink ought to be the same specification.
    RDF is really just a fancy (and important!) way of annotating links.


-- 
Liam Quin, GroveWare Inc., Toronto;  The barefoot agitator
l i a m q u i n     at    i n t e r l o g    dot   c o m
[no web page, Interlog deleted it by mistake]


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list