[
Lists Home |
Date Index |
Thread Index
]
On Thu, 27 Feb 2003 15:40:54 +0100, Robin Berjon <robin.berjon@expway.fr>
wrote:
> Elliotte Rusty Harold wrote:
>> It strikes me that a lot of people are using BinXML to mean whatever
>> they want it to mean.
>
> It's a constellation of technologies that can have very different
> features,
Exactly. That may be synonymous with "whatever they want it to mean", if
that reduces the confusion so that we can agree on what we disagree over.
The common strands in the "alternative syntaxes for XML infosets" threads
("binary" is just one particular aspect) seem to be:
- "XML" is more than just UnicodeWithAngleBrackets, it's a constellation of
technologies that are inter-related in a mutually-supporting but formally
ambiguous way. XML's success in the real world comes from their synergy,
not any one (e.g. the syntax) in isolation.
- XML 1.0 explicitly states that it should be usable for a wide variety of
applications. That is happening, probably more than anyone anticipated 5
years ago. These diverse applications developers are facing a diverse set
of challenges in dealing with the XML syntax, and are looking for ways to
get the benefits of "XML" while relaxing the constraints of XML syntax.
Among these challenges -- which are different, and almost certainly have no
common solution -- are large data volumes that are a problem when bandwidth
is constrained, slow parsing speed which is a problem in small devices or
performance-critical environments, various inefficiencies when "attaching"
binary images, executables, etc. to XML documents/messages, difficulties
with user acceptance or tools/training when creating the syntax, challenges
preserving data type information when using XML to exchange information
from one strongly typed application to another, and probably many more.
- As XML's popularity increases, the need for native but efficient XML
processing is being driven deeper and deeper into a wide variety of
applications. Proprietary solutions (e.g. binary infoset serializations)
are being designed and employed, but are being seen as less and less viable
because the logic of open standards creates demands by customers and
partners that proprietary solutions can't address.
- Therefore, there is increasing interest in making sure that the core W3C
architectures do NOT enshrine the theory that "XML" is just a syntax, but
is in fact an intertwined network of technologies built on the XML syntax,
the abstract data model that well-formed documents imply, the more concrete
data model implied by the XSDL datatypes, the declarative processing models
such as XPath, XSLT, and XQuery that build on a conception of the data
model rather than the syntax, and so on. This will probably require some
formalization of fuzzy concepts and some refactoring of existing specs in
improve their layering and modularity, i.e., so that universally applicable
concepts are in the "core", and specialized concepts are in the
"periphery." If we've learned anything from the first 5 years of XML, it's
that it can't succeed as a "big ball of mud" that pollutes document
applications with data-specific concepts or data applications with
document-specific concepts.
- Actual standardization of various alternative serializations is something
to discuss and explore at this point, not to move rapidly ahead with. Much
more real performance data, clearer delineation of communities with similar
requirements, etc., and experimentation with real alternatives is needed
first.
OK, would anyone who has spoken up in favor of "binary XML" disagree with
any of that? And what specific parts of this summary would those who have
spoken out against its ideas care to challenge?
|
