Re: Infoset as subset of useful

[John Cowan <cowan@mercury.ccil.org>]

David Brownell scripsit:
> And John Cowan's comment explains why:
> 
> > But the Infoset is not founded on any real principle at all: its
> > design is empirical.
> 
> There ought to be some organizing principle behind it.  That'd
> be better support for principled improvements/objections/etc ...
> What I really wanted out of infoset was such a principled data
> model, not a grab-bag, to support what Tim Bray described:

Very good.  You invent the principles, I'll design the data model,
and we can call it "Infoset, Part 2".

> Though it doesn't.  For example, it doesn't expose declarations
> for parsed entities (as exposed by DOM L1 and SAX2), for
> elements, or for attributes; or <![CDATA[ ... ]]> info.

Either DOM's way or the highway, and we chose the highway.
So it goes.

(This is not a criticism of the DOM WG, which had their own
special constraints to satisfy.)

> I suspect that we're probably not going to get a more useful
> formal XML data model for maintaining such consistency
> than what Infoset delivers.  It's better than the XML 1.0 spec
> for that purpose, even with its current omissions.

Yup.  We are, after all, dealing with SGML, devised by a lawyer
from a common-law (i.e. explicitly anti-theoretical) tradition.

-- 
John Cowan                                   cowan@ccil.org
One art/there is/no less/no more/All things/to do/with sparks/galore
	--Douglas Hofstadter