[
Lists Home |
Date Index |
Thread Index
]
"Alaric B. Snell" wrote:
> Thing is, I reckon any reasonably binxml parser would come with a tool (and it
> would be a simple tool to write) that maps it to and from textual XML. And it
> should be the case that parsers accept both text-XML and bin-XML while
> outputting your choice of format
Petitio principii, I'm afraid. An XML parser, by definition, accepts XML as
syntax as input. If there is *a* (single) mapping known to a parser from XML
syntax to a particular binary form, then that parser is suited only for one
particular and limited application of XML [see last week's iteration of that
permathread]. XML says nothing of the particular instantiation which is built on
the output of a parse; if it does it is not XML but one narrowly constrained
application of a particular datastructure, and that of course requires a priori
agreement between creators and all of the consumers of their documents as to the
specific schematics of that instantiated structure.
Again, as I have often pointed out, this is not a new controversy. Some of us who
have been through it before came to XML because the well-formed document was the
essential premise to avoid the pernicious path of infosets and the fluid
preverbal Gestalt. If you premise some new 'XML' on a datastructure rather than
on the syntax of the concrete instance it becomes useless for the very jobs that
it is now uniquely suited for. Fortunately, there are now alternative tools. LMNL
in particular, though it is defined in terms of a data model, insists on the
parsing of a text as the basis of constructing ranges against which a variety of
data structures may be instantiated. This has the salutary effect of allowing
LMNL to handle a number of syntaxes mapped concurrently to differing
datastructures (not so incidentally trumping on both sides your mapping of a
single syntax to a single binXML).
Respectfully,
Walter Perry
|
