RE: !BEHAVIOR

Andrew Layman wrote:
	I disagree with the recent suggestions to add BEHAVIOR
attributes or
	processing instructions to data. (And the fact that XML-LINK
mixes these
	two reflects insufficient layering in the XML-LINK spec.)


Is the bit between parentheses part of what Andrew _disagrees_ with, or
is it a statement he is making to underline his disagreement with the
idea of  BEHAVIOR attributes for data elements?

It might help if we separate out some of the issues which have been
raised.

1)	I argued in an earlier posting ("Structure and Behaviour -
Formatting and Behaviour") that behavioural attributes (ACTUATE and
SHOW) should be removed from XML-LINK. What they specify should not be
tied to the element, but should be specified in a stylesheet. No-one has
so far argued against this on this list, as far as I am aware.

2)	I also argued that stylesheets should be able to provide
information as to how the data should be processed and/or presented,
including aspects of processing and behaviour which go beyond just the
formatting of a two-dimensional display. In other words, there is no
logical difference between "formatting" and "behaviour". A stylesheet
syntax should be rich enough to specify whatever kind of formatting
and/or behaviour is needed, depending on the nature and purpose of the
application which uses it. Here too, I am not aware of anyone having
objected to this point of view.

3)	Martin Bryan suggested that a BEHAVIOR attribute should be
available on XML elements (link elements or not) to point to a local
processor, but with a mechanism for indirection. This attribute could be
used to specify which stylesheet should be applied to a given element.
Presumably this would allow different stylesheets to be applied
different elements (even elements of the same type) within the same
document. This raises some interesting issues, such as: "Can the user
override the stylesheet specified in the BEHAVIOR attribute?", "Can
meta-data within the document (syntax yet to be decided) be used to
specify a stylesheet to be applied to the whole document?", "If one
stylesheet is specified for the document as a whole, another for a
particular element, and a third chosen by the user for a specific
purpose on a specific occasion, which overrides which?" Are the
complexities of this kind of scenario such as to argue against the use
of BEHAVIOR attributes on individual elements? 

There has been some interesting discussion on the possible overlap
between issues of semantics, behaviour, namespaces and notations. This
is all important and useful. I'm looking forward to seeing how the
discussion plays out.

Andrew's comment above brings us back to the vital issue of indirection
(layering). Let us not forget that the separation of formatting from
content is arguably one of SGML's greatest achievements. Unfortunately,
SGML in 1986 did not provide a standard way of doing formatting - yet
formatting has to be done. In the absence of a formatting syntax in SGML
(before the arrival of DSSSL) people were naturally tempted to use the
DTD to specify formatting. Hence the CALS table model, for example.
Hence HTML, which uses primarily format-oriented tags. 

Let us not fall back into the same trap by having attributes in XML-LINK
which suggest that behaviour (presentation) is an attribute of the link
itself. Let us get rid of ACUATE and SHOW, and if we do add a BEHAVIOR
attribute, let us make the indirection mechanism very very clear.


	-----Original Message-----
	From:	Andrew Layman [SMTP:andrewl@microsoft.com]
	Sent:	Tuesday, June 10, 1997 4:41 PM
	To:	XML Working Group (E-mail)
	Subject:	!BEHAVIOR

	--Andrew Layman
	   AndrewL@microsoft.com

Received on Wednesday, 11 June 1997 13:03:22 UTC